feat(application): PSD layer separation and image quality/count threading

Convert GPT Image 2 layered PSD output into canvas image/text nodes with a local separation fallback (new layer_separation.go, oov/psd dependency), default layer-separation/edit-elements output_format to psd, and thread imageQuality/imageCount through project generation and job processing. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
2026-07-09 20:42:42 +08:00
parent 00b70298f2
commit cb24ffd185
7 changed files with 565 additions and 78 deletions
@@ -311,6 +311,98 @@ func TestLayerSeparationPosterKeepsComplexBackgroundAndExtractsSmallElements(t *
 	}
 }

+func TestLayerSeparationDocumentPlacesSmallImagesAndTextOverBackground(t *testing.T) {
+	ctx := context.Background()
+	service := NewDesignService(nil, nil, nil, nil, nil)
+	source := testPosterSourceImage()
+	background := testPosterCleanBackgroundImage()
+	ornament := image.NewNRGBA(image.Rect(54, 44, 66, 56))
+	for y := 44; y < 56; y++ {
+		for x := 54; x < 66; x++ {
+			ornament.SetNRGBA(x, y, color.NRGBA{R: 225, G: 32, B: 44, A: 255})
+		}
+	}
+	textRaster := image.NewNRGBA(image.Rect(8, 8, 36, 18))
+	for y := 8; y < 18; y++ {
+		for x := 8; x < 36; x++ {
+			textRaster.SetNRGBA(x, y, color.NRGBA{R: 17, G: 17, B: 17, A: 255})
+		}
+	}
+
+	result, err := service.layerSeparationResultFromDocument(ctx, "project-layered-doc", source, layerSeparationDocument{
+		Bounds: source.Bounds(),
+		Layers: []layerSeparationDocumentLayer{
+			{Name: "clean background", Rect: source.Bounds(), Image: background, Opacity: 255},
+			{Name: "decorative element", Rect: ornament.Bounds(), Image: ornament, Opacity: 255},
+			{Name: "text title", Rect: textRaster.Bounds(), Image: textRaster, Opacity: 255},
+		},
+	}, design.TextExtraction{
+		Layers: []design.ExtractedTextLayer{{Text: "SALE", X: 0.1, Y: 0.1, Width: 0.35, Height: 0.125, Color: "#111111"}},
+	})
+	if err != nil {
+		t.Fatal(err)
+	}
+	if len(result.Foregrounds) != 1 {
+		t.Fatalf("expected one small image layer, got %#v", result.Foregrounds)
+	}
+	if got := result.Foregrounds[0].BBox; len(got) != 4 || got[0] != 54 || got[1] != 44 || got[2] != 66 || got[3] != 56 {
+		t.Fatalf("unexpected foreground bbox: %#v", got)
+	}
+	if len(result.TextExtraction.Layers) != 1 || result.TextExtraction.Layers[0].Text != "SALE" {
+		t.Fatalf("expected editable OCR text, got %#v", result.TextExtraction.Layers)
+	}
+	backgroundData, _, err := loadImageContent(ctx, result.BackgroundURL)
+	if err != nil {
+		t.Fatal(err)
+	}
+	decodedBackground, _, err := image.Decode(bytes.NewReader(backgroundData))
+	if err != nil {
+		t.Fatal(err)
+	}
+	ornamentPixel := color.NRGBAModel.Convert(decodedBackground.At(59, 49)).(color.NRGBA)
+	if ornamentPixel.R > 200 && ornamentPixel.G < 80 {
+		t.Fatalf("expected foreground element removed from background, got %#v", ornamentPixel)
+	}
+}
+
+func TestLayerSeparationDocumentAllowsLargeSubjectOnPlainBackground(t *testing.T) {
+	ctx := context.Background()
+	service := NewDesignService(nil, nil, nil, nil, nil)
+	bounds := image.Rect(0, 0, 100, 100)
+	source := image.NewNRGBA(bounds)
+	background := image.NewNRGBA(bounds)
+	for y := 0; y < 100; y++ {
+		for x := 0; x < 100; x++ {
+			source.SetNRGBA(x, y, color.NRGBA{R: 230, G: 228, B: 224, A: 255})
+			background.SetNRGBA(x, y, color.NRGBA{R: 230, G: 228, B: 224, A: 255})
+		}
+	}
+	subject := image.NewNRGBA(image.Rect(28, 14, 74, 92))
+	for y := 14; y < 92; y++ {
+		for x := 28; x < 74; x++ {
+			subject.SetNRGBA(x, y, color.NRGBA{R: 36, G: 28, B: 22, A: 255})
+			source.SetNRGBA(x, y, color.NRGBA{R: 36, G: 28, B: 22, A: 255})
+		}
+	}
+
+	result, err := service.layerSeparationResultFromDocument(ctx, "project-plain-subject", source, layerSeparationDocument{
+		Bounds: bounds,
+		Layers: []layerSeparationDocumentLayer{
+			{Name: "background", Rect: bounds, Image: background, Opacity: 255},
+			{Name: "subject", Rect: subject.Bounds(), Image: subject, Opacity: 255},
+		},
+	}, design.TextExtraction{})
+	if err != nil {
+		t.Fatal(err)
+	}
+	if len(result.Foregrounds) != 1 {
+		t.Fatalf("expected large subject foreground on plain background, got %#v", result.Foregrounds)
+	}
+	if got := result.Foregrounds[0].BBox; len(got) != 4 || got[0] != 28 || got[1] != 14 || got[2] != 74 || got[3] != 92 {
+		t.Fatalf("unexpected subject bbox: %#v", got)
+	}
+}
+
 func waitForStandaloneGeneratorTask(t *testing.T, service *DesignService, taskID string) design.GeneratorTask {
 	t.Helper()
 	deadline := time.Now().Add(2 * time.Second)
@@ -360,6 +452,10 @@ func testLayerSeparationForegroundPNG(t *testing.T) []byte {
 }

 func testPosterSourceDataURL() string {
+	return testPNGDataURLFromImage(testPosterSourceImage())
+}
+
+func testPosterSourceImage() *image.NRGBA {
 	img := image.NewNRGBA(image.Rect(0, 0, 80, 80))
 	for y := 0; y < 80; y++ {
 		for x := 0; x < 80; x++ {
@@ -381,7 +477,22 @@ func testPosterSourceDataURL() string {
 			img.SetNRGBA(x, y, color.NRGBA{R: 225, G: 32, B: 44, A: 255})
 		}
 	}
-	return testPNGDataURLFromImage(img)
+	return img
+}
+
+func testPosterCleanBackgroundImage() *image.NRGBA {
+	img := image.NewNRGBA(image.Rect(0, 0, 80, 80))
+	for y := 0; y < 80; y++ {
+		for x := 0; x < 80; x++ {
+			img.SetNRGBA(x, y, color.NRGBA{R: uint8(190 + x/3), G: uint8(186 + y/4), B: uint8(174 + (x+y)/8), A: 255})
+		}
+	}
+	for y := 38; y < 70; y++ {
+		for x := 8; x < 50; x++ {
+			img.SetNRGBA(x, y, color.NRGBA{R: 64, G: 106, B: 204, A: 255})
+		}
+	}
+	return img
 }

 func testPosterForegroundPNG(t *testing.T) []byte {