LibGfx/JBIG2+Tests: Add support for refining page contents

Tests/LibGfx/TestImageDecoder.cpp

@@ -366,6 +366,8 @@ TEST_CASE(test_jbig2_decode)
         TEST_INPUT("jbig2/bitmap-trailing-7fff-stripped-harder.jbig2"sv),
         TEST_INPUT("jbig2/bitmap-trailing-7fff-stripped-harder-refine.jbig2"sv),
         TEST_INPUT("jbig2/bitmap-refine.jbig2"sv),
+        TEST_INPUT("jbig2/bitmap-refine-page.jbig2"sv),
+        TEST_INPUT("jbig2/bitmap-refine-page-subrect.jbig2"sv),
         TEST_INPUT("jbig2/bitmap-refine-customat.jbig2"sv),
         TEST_INPUT("jbig2/bitmap-refine-lossless.jbig2"sv),
         TEST_INPUT("jbig2/bitmap-refine-refine.jbig2"sv),
@@ -420,7 +422,6 @@ TEST_CASE(test_jbig2_decode)
         // - intermediate direct regions (code support added in #26197)
         // - symbol refinement referring to symbol in same segment
         // Missing tests for things that aren't implemented yet:
-        // - immediate refinement regions not referring to a direct region (i.e. refining the page)
         // - exttemplate
         // - colors
     };

Tests/LibGfx/test-inputs/jbig2/json/bitmap-refine-page-subrect.json

@@ -0,0 +1,60 @@
+{
+    "global_header": {
+        "organization": "sequential",
+        "number_of_pages": 1
+    },
+    "segments": [
+        {
+            "segment_number": 0,
+            "type": "page_information",
+            "page_association": 1,
+            "data": {
+                "page_width": 399,
+                "page_height": 400,
+                "flags": {
+                    "direct_region_segments_override_default_combination_operator": true
+                }
+            }
+        },
+        {
+            "segment_number": 1,
+            "type": "generic_region",
+            "page_association": 1,
+            "data": {
+                "image_data": {
+                    "from_file": "bitmap-blemish.bmp"
+                }
+            }
+        },
+        {
+            "segment_number": 2,
+            "type": "generic_refinement_region",
+            "page_association": 1,
+            "data": {
+                "region_segment_information": {
+                    "x": 10,
+                    "y": 20,
+                    "width": 110,
+                    "height": 380,
+                    "flags": {
+                        "external_combination_operator": "replace"
+                    }
+                },
+                "image_data": {
+                    "from_file": "bitmap.bmp",
+                    "crop": {
+                        "x": 10,
+                        "y": 20,
+                        "width": 110,
+                        "height": 380
+                    }
+                }
+            }
+        },
+        {
+            "segment_number": 3,
+            "type": "end_of_page",
+            "page_association": 1
+        }
+    ]
+}

Tests/LibGfx/test-inputs/jbig2/json/bitmap-refine-page.json

@@ -0,0 +1,50 @@
+{
+    "global_header": {
+        "organization": "sequential",
+        "number_of_pages": 1
+    },
+    "segments": [
+        {
+            "segment_number": 0,
+            "type": "page_information",
+            "page_association": 1,
+            "data": {
+                "page_width": 399,
+                "page_height": 400,
+                "flags": {
+                    "direct_region_segments_override_default_combination_operator": true
+                }
+            }
+        },
+        {
+            "segment_number": 1,
+            "type": "generic_region",
+            "page_association": 1,
+            "data": {
+                "image_data": {
+                    "from_file": "bitmap-blemish.bmp"
+                }
+            }
+        },
+        {
+            "segment_number": 2,
+            "type": "generic_refinement_region",
+            "page_association": 1,
+            "data": {
+                "region_segment_information": {
+                    "flags": {
+                        "external_combination_operator": "replace"
+                    }
+                },
+                "image_data": {
+                    "from_file": "bitmap.bmp"
+                }
+            }
+        },
+        {
+            "segment_number": 3,
+            "type": "end_of_page",
+            "page_association": 1
+        }
+    ]
+}

Userland/Libraries/LibGfx/ImageFormats/JBIG2Loader.cpp

@@ -3190,16 +3190,19 @@ static ErrorOr<RegionResult> decode_generic_refinement_region(JBIG2LoadingContex
 
     // "3) Determine the buffer associated with the region segment that this segment refers to."
     // Details described in 7.4.7.4 Reference bitmap selection.
-    BilevelImage const* reference_bitmap = nullptr;
-    if (segment.referred_to_segments.size() == 1) {
-        reference_bitmap = segment.referred_to_segments[0]->aux_buffer.ptr();
-        VERIFY(reference_bitmap->width() == segment.referred_to_segments[0]->aux_buffer_information_field.width);
-        VERIFY(reference_bitmap->height() == segment.referred_to_segments[0]->aux_buffer_information_field.height);
-    } else {
-        // When adding support for this and for intermediate generic refinement regions, make sure to only allow
-        // this case for immediate generic refinement regions.
-        return Error::from_string_literal("JBIG2ImageDecoderPlugin: Generic refinement region without reference segment not yet implemented");
-    }
+    BilevelSubImage reference_bitmap = [&]() {
+        if (segment.referred_to_segments.size() == 1) {
+            auto reference_bitmap = segment.referred_to_segments[0]->aux_buffer;
+            VERIFY(reference_bitmap->width() == segment.referred_to_segments[0]->aux_buffer_information_field.width);
+            VERIFY(reference_bitmap->height() == segment.referred_to_segments[0]->aux_buffer_information_field.height);
+            return reference_bitmap->as_subbitmap();
+        }
+
+        // Enforced by validate_segment_header_references() earlier.
+        VERIFY(segment.type() != JBIG2::SegmentType::IntermediateGenericRefinementRegion);
+
+        return context.page.bits->subbitmap(information_field.rect());
+    }();
 
     // "4) Invoke the generic refinement region decoding procedure described in 6.3, with the parameters to the
     //     generic refinement region decoding procedure set as shown in Table 38."
@@ -3208,8 +3211,7 @@ static ErrorOr<RegionResult> decode_generic_refinement_region(JBIG2LoadingContex
     inputs.region_width = information_field.width;
     inputs.region_height = information_field.height;
     inputs.gr_template = arithmetic_coding_template;
-    auto subbitmap = reference_bitmap->as_subbitmap();
-    inputs.reference_bitmap = &subbitmap;
+    inputs.reference_bitmap = &reference_bitmap;
     inputs.reference_x_offset = 0;
     inputs.reference_y_offset = 0;
     inputs.is_typical_prediction_used = typical_prediction_generic_refinement_on;
@@ -3666,7 +3668,7 @@ ErrorOr<ImageFrameDescriptor> JBIG2ImageDecoderPlugin::frame(size_t index, Optio
     return ImageFrameDescriptor { move(bitmap), 0 };
 }
 
-ErrorOr<ByteBuffer> JBIG2ImageDecoderPlugin::decode_embedded(Vector<ReadonlyBytes> data)
+ErrorOr<NonnullRefPtr<BilevelImage>> JBIG2ImageDecoderPlugin::decode_embedded(Vector<ReadonlyBytes> data)
 {
     auto plugin = TRY(adopt_nonnull_own_or_enomem(new (nothrow) JBIG2ImageDecoderPlugin({})));
     plugin->m_context->organization = JBIG2::Organization::Embedded;
@@ -3683,7 +3685,7 @@ ErrorOr<ByteBuffer> JBIG2ImageDecoderPlugin::decode_embedded(Vector<ReadonlyByte
 
     TRY(decode_data(*plugin->m_context));
 
-    return plugin->m_context->page.bits->to_byte_buffer();
+    return *plugin->m_context->page.bits;
 }
 
 }

Userland/Libraries/LibGfx/ImageFormats/JBIG2Loader.h

@@ -12,6 +12,7 @@
 
 namespace Gfx {
 
+class BilevelImage;
 struct JBIG2LoadingContext;
 struct MQArithmeticCoderContext;
 class MQArithmeticDecoder;
@@ -69,7 +70,7 @@ class JBIG2ImageDecoderPlugin : public ImageDecoderPlugin {
     virtual size_t frame_count() override;
     virtual ErrorOr<ImageFrameDescriptor> frame(size_t index, Optional<IntSize> ideal_size = {}) override;
 
-    static ErrorOr<ByteBuffer> decode_embedded(Vector<ReadonlyBytes>);
+    static ErrorOr<NonnullRefPtr<BilevelImage>> decode_embedded(Vector<ReadonlyBytes>);
 
 private:
     JBIG2ImageDecoderPlugin(JBIG2DecoderOptions);

Userland/Libraries/LibGfx/ImageFormats/JBIG2Shared.h

@@ -11,6 +11,7 @@
 #include <AK/Types.h>
 #include <AK/Vector.h>
 #include <LibGfx/ImageFormats/MQArithmeticCoder.h>
+#include <LibGfx/Rect.h>
 
 namespace Gfx::JBIG2 {
 
@@ -118,6 +119,11 @@ struct [[gnu::packed]] RegionSegmentInformationField {
     BigEndian<u32> y_location;
     u8 flags { 0 };
 
+    IntRect rect() const
+    {
+        return { x_location, y_location, width, height };
+    }
+
     CombinationOperator external_combination_operator() const
     {
         VERIFY((flags & 0x7) <= 4);

Userland/Libraries/LibGfx/ImageFormats/JBIG2Writer.cpp

@@ -21,6 +21,7 @@
 #include <LibGfx/Bitmap.h>
 #include <LibGfx/ImageFormats/BilevelImage.h>
 #include <LibGfx/ImageFormats/CCITTEncoder.h>
+#include <LibGfx/ImageFormats/JBIG2Loader.h>
 #include <LibGfx/ImageFormats/JBIG2Writer.h>
 #include <LibGfx/ImageFormats/MQArithmeticCoder.h>
 #include <LibTextCodec/Encoder.h>
@@ -337,7 +338,7 @@ namespace {
 struct GenericRefinementRegionEncodingInputParameters {
     BilevelImage const& image;                                          // Of dimensions "GRW" x "GRH" in spec terms.
     u8 gr_template { 0 };                                               // "GRTEMPLATE" in spec.
-    BilevelImage const* reference_bitmap { nullptr };                   // "GRREFERENCE" in spec.
+    BilevelSubImage reference_bitmap;                                   // "GRREFERENCE" in spec.
     i32 reference_x_offset { 0 };                                       // "GRREFERENCEDX" in spec.
     i32 reference_y_offset { 0 };                                       // "GRREFERENCEDY" in spec.
     bool is_typical_prediction_used { false };                          // "TPGRON" in spec.
@@ -378,7 +379,7 @@ static ErrorOr<void> generic_refinement_region_encoding_procedure(GenericRefinem
     };
 
     // Figure 12 – 13-pixel refinement template showing the AT pixels at their nominal locations
-    constexpr auto compute_context_0 = [](ReadonlySpan<JBIG2::AdaptiveTemplatePixel> adaptive_pixels, BilevelImage const& reference, int reference_x, int reference_y, BilevelImage const& buffer, int x, int y) -> u16 {
+    constexpr auto compute_context_0 = [](ReadonlySpan<JBIG2::AdaptiveTemplatePixel> adaptive_pixels, BilevelSubImage const& reference, int reference_x, int reference_y, BilevelImage const& buffer, int x, int y) -> u16 {
         u16 result = 0;
 
         for (int dy = -1; dy <= 1; ++dy) {
@@ -399,7 +400,7 @@ static ErrorOr<void> generic_refinement_region_encoding_procedure(GenericRefinem
     };
 
     // Figure 13 – 10-pixel refinement template
-    constexpr auto compute_context_1 = [](ReadonlySpan<JBIG2::AdaptiveTemplatePixel>, BilevelImage const& reference, int reference_x, int reference_y, BilevelImage const& buffer, int x, int y) -> u16 {
+    constexpr auto compute_context_1 = [](ReadonlySpan<JBIG2::AdaptiveTemplatePixel>, BilevelSubImage const& reference, int reference_x, int reference_y, BilevelImage const& buffer, int x, int y) -> u16 {
         u16 result = 0;
 
         for (int dy = -1; dy <= 1; ++dy) {
@@ -440,10 +441,10 @@ static ErrorOr<void> generic_refinement_region_encoding_procedure(GenericRefinem
         auto predict = [&](size_t x, size_t y) -> Optional<bool> {
             // "• a 3 × 3 pixel array in the reference bitmap (Figure 16), centred at the location
             //    corresponding to the current pixel, contains pixels all of the same value."
-            bool prediction = get_pixel(*inputs.reference_bitmap, x - inputs.reference_x_offset - 1, y - inputs.reference_y_offset - 1);
+            bool prediction = get_pixel(inputs.reference_bitmap, x - inputs.reference_x_offset - 1, y - inputs.reference_y_offset - 1);
             for (int dy = -1; dy <= 1; ++dy)
                 for (int dx = -1; dx <= 1; ++dx)
-                    if (get_pixel(*inputs.reference_bitmap, x - inputs.reference_x_offset + dx, y - inputs.reference_y_offset + dy) != prediction)
+                    if (get_pixel(inputs.reference_bitmap, x - inputs.reference_x_offset + dx, y - inputs.reference_y_offset + dy) != prediction)
                         return {};
             return prediction;
         };
@@ -471,7 +472,7 @@ static ErrorOr<void> generic_refinement_region_encoding_procedure(GenericRefinem
             // "c) If LTP = 0 then, from left to right, explicitly decode all pixels of the current row of GRREG. The
             //     procedure for each pixel is as follows:"
             for (size_t x = 0; x < width; ++x) {
-                u16 context = compute_context(inputs.adaptive_template_pixels, *inputs.reference_bitmap, x - inputs.reference_x_offset, y - inputs.reference_y_offset, inputs.image, x, y);
+                u16 context = compute_context(inputs.adaptive_template_pixels, inputs.reference_bitmap, x - inputs.reference_x_offset, y - inputs.reference_y_offset, inputs.image, x, y);
                 encoder.encode_bit(inputs.image.get_bit(x, y), contexts.contexts[context]);
             }
         } else {
@@ -484,7 +485,7 @@ static ErrorOr<void> generic_refinement_region_encoding_procedure(GenericRefinem
                 // TPGRON must be 1 if LTP is set. (The spec has an explicit "TPGRON is 1 AND" check here, but it is pointless.)
                 VERIFY(inputs.is_typical_prediction_used);
                 if (!prediction.has_value()) {
-                    u16 context = compute_context(inputs.adaptive_template_pixels, *inputs.reference_bitmap, x - inputs.reference_x_offset, y - inputs.reference_y_offset, inputs.image, x, y);
+                    u16 context = compute_context(inputs.adaptive_template_pixels, inputs.reference_bitmap, x - inputs.reference_x_offset, y - inputs.reference_y_offset, inputs.image, x, y);
                     encoder.encode_bit(inputs.image.get_bit(x, y), contexts.contexts[context]);
                 }
             }
@@ -726,7 +727,7 @@ static ErrorOr<void> text_region_encoding_procedure(TextRegionEncodingInputParam
         auto reference_bitmap = TRY(symbol_image(symbol));
         GenericRefinementRegionEncodingInputParameters refinement_inputs {
             .image = symbol_instance.refinement_data->refines_to,
-            .reference_bitmap = reference_bitmap,
+            .reference_bitmap = reference_bitmap->as_subbitmap(),
         };
 
         // FIXME: Instead, just compute the delta here instead of having it be passed in?
@@ -1136,7 +1137,7 @@ static ErrorOr<ByteBuffer> symbol_dictionary_encoding_procedure(SymbolDictionary
         // Table 18 – Parameters used to decode a symbol's bitmap when REFAGGNINST = 1
         GenericRefinementRegionEncodingInputParameters refinement_inputs {
             .image = *refinement_image.refines_to,
-            .reference_bitmap = IBO,
+            .reference_bitmap = IBO->as_subbitmap(),
         };
         refinement_inputs.gr_template = inputs.refinement_template;
         refinement_inputs.reference_x_offset = refinement_image.delta_x_offset;
@@ -1543,6 +1544,13 @@ struct SerializedSegmentData {
 };
 
 struct JBIG2EncodingContext {
+    JBIG2EncodingContext(Vector<JBIG2::SegmentData> const& segments)
+        : segments(segments)
+    {
+    }
+
+    Vector<JBIG2::SegmentData> const& segments;
+
     HashMap<u32, JBIG2::SegmentData const*> segment_by_id;
 
     HashMap<u32, SerializedSegmentData> segment_data_by_id;
@@ -2266,24 +2274,46 @@ static ErrorOr<void> encode_generic_refinement_region(JBIG2::GenericRefinementRe
     // 7.4.7 Generic refinement region syntax
     if (header.referred_to_segments.size() > 1)
         return Error::from_string_literal("JBIG2Writer: Generic refinement region must refer to at most one segment");
-    if (header.referred_to_segments.size() == 0)
-        return Error::from_string_literal("JBIG2Writer: Generic refinement region refining page not yet implemented");
 
-    auto maybe_segment = context.segment_by_id.get(header.referred_to_segments[0].segment_number);
-    if (!maybe_segment.has_value())
-        return Error::from_string_literal("JBIG2Writer: Could not find referred-to segment for generic refinement region");
-    auto const& referred_to_segment = *maybe_segment.value();
+    // 7.4.7.4 Reference bitmap selection
+    auto const reference_bitmap = TRY([&]() -> ErrorOr<BilevelSubImage> {
+        // "If this segment refers to another region segment, then set the reference bitmap GRREFERENCE to be the current
+        //  contents of the auxiliary buffer associated with the region segment that this segment refers to."
+        if (header.referred_to_segments.size() == 1) {
+            auto maybe_segment = context.segment_by_id.get(header.referred_to_segments[0].segment_number);
+            if (!maybe_segment.has_value())
+                return Error::from_string_literal("JBIG2Writer: Could not find referred-to segment for generic refinement region");
+            auto const& referred_to_segment = *maybe_segment.value();
+
+            return TRY(referred_to_segment.data.visit(
+                           [](JBIG2::IntermediateGenericRegionSegmentData const& generic_region_wrapper) -> ErrorOr<BilevelImage const*> {
+                               return generic_region_wrapper.generic_region.image;
+                           },
+                           [](JBIG2::IntermediateGenericRefinementRegionSegmentData const& generic_refinement_region_wrapper) -> ErrorOr<BilevelImage const*> {
+                               return generic_refinement_region_wrapper.generic_refinement_region.image;
+                           },
+                           [](auto const&) -> ErrorOr<BilevelImage const*> {
+                               return Error::from_string_literal("JBIG2Writer: Generic refinement region can only refer to intermediate region segments");
+                           }))
+                ->as_subbitmap();
+        }
 
-    auto const* reference_bitmap = TRY(referred_to_segment.data.visit(
-        [](JBIG2::IntermediateGenericRegionSegmentData const& generic_region_wrapper) -> ErrorOr<BilevelImage const*> {
-            return generic_region_wrapper.generic_region.image;
-        },
-        [](JBIG2::IntermediateGenericRefinementRegionSegmentData const& generic_refinement_region_wrapper) -> ErrorOr<BilevelImage const*> {
-            return generic_refinement_region_wrapper.generic_refinement_region.image;
-        },
-        [](auto const&) -> ErrorOr<BilevelImage const*> {
-            return Error::from_string_literal("JBIG2Writer: Generic refinement region can only refer to intermediate region segments");
-        }));
+        // "If this segment does not refer to another region segment, set GRREFERENCE to be a bitmap containing the current
+        //  contents of the page buffer (see clause 8), restricted to the area of the page buffer specified by this segment's region
+        //  segment information field."
+        VERIFY(header.referred_to_segments.size() == 0);
+        Vector<ReadonlyBytes> preceding_segments_on_same_page;
+        for (auto const& segment : context.segments) {
+            if (segment.header.page_association == 0 || segment.header.page_association == header.page_association) {
+                if (&segment.header == &header)
+                    break;
+                auto const& data = context.segment_data_by_id.get(segment.header.segment_number);
+                preceding_segments_on_same_page.append(data->data);
+            }
+        }
+        auto bitmap = TRY(JBIG2ImageDecoderPlugin::decode_embedded(preceding_segments_on_same_page));
+        return bitmap->subbitmap(generic_refinement_region.region_segment_information.rect());
+    }());
 
     GenericRefinementRegionEncodingInputParameters inputs {
         .image = *generic_refinement_region.image,
@@ -2624,7 +2654,7 @@ ErrorOr<void> JBIG2Writer::encode_with_explicit_data(Stream& stream, JBIG2::File
 
     TRY(encode_jbig2_header(stream, file_data.header));
 
-    JBIG2EncodingContext context;
+    JBIG2EncodingContext context { file_data.segments };
     for (auto const& segment : file_data.segments) {
         if (context.segment_by_id.set(segment.header.segment_number, &segment) != HashSetResult::InsertedNewEntry)
             return Error::from_string_literal("JBIG2Writer: Duplicate segment number");