Cthulhu is no longer a fork. Now it is a screen reader written in Golang.

internal/ocr/preprocess.go

@@ -0,0 +1,57 @@
+// Copyright (c) 2026 Storm Dragon
+// Authors: Storm Dragon, Angus Kola
+// SPDX-License-Identifier: GPL-3.0-or-later
+
+package ocr
+
+import (
+	"context"
+	"image"
+	"image/color"
+)
+
+// Preprocess returns owned pixels in recognition coordinates. Scale uses nearest
+// neighbor sampling so recognized word boxes map back through the capture transform.
+func Preprocess(ctx context.Context, source image.Image, options Options) (image.Image, error) {
+	if source == nil || !valid_dimensions(source.Bounds().Dx(), source.Bounds().Dy()) {
+		return nil, ErrInvalidImage
+	}
+	scale := options.Scale
+	if scale == 0 {
+		scale = 1
+	}
+	width, height := source.Bounds().Dx(), source.Bounds().Dy()
+	if scale < 1 || scale > 4 || height > MaxPixels/scale || !valid_dimensions(width*scale, height*scale) ||
+		options.Threshold < 0 || options.Threshold > 255 {
+		return nil, ErrLimit
+	}
+	if scale == 1 && !options.Grayscale && !options.Invert && !options.BlackWhite {
+		return source, nil
+	}
+	processed := image.NewRGBA(image.Rect(0, 0, width*scale, height*scale))
+	for y := 0; y < height*scale; y++ {
+		if y%32 == 0 && ctx.Err() != nil {
+			return nil, ctx.Err()
+		}
+		for x := 0; x < width*scale; x++ {
+			r, g, b, _ := source.At(source.Bounds().Min.X+x/scale, source.Bounds().Min.Y+y/scale).RGBA()
+			red, green, blue := uint8(r>>8), uint8(g>>8), uint8(b>>8)
+			gray := uint8((299*uint32(red) + 587*uint32(green) + 114*uint32(blue) + 500) / 1000)
+			if options.Grayscale || options.BlackWhite {
+				red, green, blue = gray, gray, gray
+			}
+			if options.Invert {
+				red, green, blue = 255-red, 255-green, 255-blue
+			}
+			if options.BlackWhite {
+				value := uint8(0)
+				if int(red) >= options.Threshold {
+					value = 255
+				}
+				red, green, blue = value, value, value
+			}
+			processed.SetRGBA(x, y, color.RGBA{R: red, G: green, B: blue, A: 255})
+		}
+	}
+	return processed, ctx.Err()
+}