poe2-bot/tools/OcrDaemon/EdgeCropHandler.cs

namespace OcrDaemon;

using System.Drawing;
using System.Drawing.Imaging;
using System.Runtime.InteropServices;

class EdgeCropHandler
{
    [StructLayout(LayoutKind.Sequential)]
    private struct POINT { public int X, Y; }

    [DllImport("user32.dll")]
    private static extern bool GetCursorPos(out POINT lpPoint);

    public (Bitmap cropped, Bitmap fullCapture, RegionRect region)? EdgeCrop(Request req, EdgeCropParams p)
    {
        int cursorX, cursorY;
        if (req.CursorX.HasValue && req.CursorY.HasValue)
        {
            cursorX = req.CursorX.Value;
            cursorY = req.CursorY.Value;
        }
        else
        {
            GetCursorPos(out var pt);
            cursorX = pt.X;
            cursorY = pt.Y;
        }

        var fullCapture = ScreenCapture.CaptureOrLoad(req.File, null);
        int w = fullCapture.Width;
        int h = fullCapture.Height;

        var bmpData = fullCapture.LockBits(new Rectangle(0, 0, w, h), ImageLockMode.ReadOnly, PixelFormat.Format32bppArgb);
        byte[] px = new byte[bmpData.Stride * h];
        Marshal.Copy(bmpData.Scan0, px, 0, px.Length);
        fullCapture.UnlockBits(bmpData);
        int stride = bmpData.Stride;

        int darkThresh = p.DarkThresh;
        int colGap = p.RunGapTolerance;
        int maxGap = p.MaxGap;

        // ── Phase 1: Per-row horizontal extent ──
        // Scan left/right from cursorX per row. Gap tolerance bridges through text.
        // Percentile-based filtering for robustness.
        int bandHalf = p.MinDarkRun; // repurpose: half-height of horizontal scan band
        int bandTop = Math.Max(0, cursorY - bandHalf);
        int bandBot = Math.Min(h - 1, cursorY + bandHalf);

        var leftExtents = new List<int>();
        var rightExtents = new List<int>();

        for (int y = bandTop; y <= bandBot; y++)
        {
            int rowOff = y * stride;
            int ci = rowOff + cursorX * 4;
            int cBright = (px[ci] + px[ci + 1] + px[ci + 2]) / 3;
            if (cBright >= darkThresh) continue;

            int leftEdge = cursorX;
            int gap = 0;
            for (int x = cursorX - 1; x >= 0; x--)
            {
                int i = rowOff + x * 4;
                int brightness = (px[i] + px[i + 1] + px[i + 2]) / 3;
                if (brightness < darkThresh) { leftEdge = x; gap = 0; }
                else if (++gap > colGap) break;
            }

            int rightEdge = cursorX;
            gap = 0;
            for (int x = cursorX + 1; x < w; x++)
            {
                int i = rowOff + x * 4;
                int brightness = (px[i] + px[i + 1] + px[i + 2]) / 3;
                if (brightness < darkThresh) { rightEdge = x; gap = 0; }
                else if (++gap > colGap) break;
            }

            leftExtents.Add(leftEdge);
            rightExtents.Add(rightEdge);
        }

        if (leftExtents.Count < 10)
        {
            Console.Error.WriteLine($"  edge-crop: too few dark rows ({leftExtents.Count})");
            fullCapture.Dispose();
            return null;
        }

        leftExtents.Sort();
        rightExtents.Sort();

        // Use RowThreshDiv/ColThreshDiv as percentile denominators
        // e.g., RowThreshDiv=4 → 25th percentile for left, ColThreshDiv=4 → 75th for right
        int leftPctIdx = leftExtents.Count / p.RowThreshDiv;
        int rightPctIdx = rightExtents.Count * (p.ColThreshDiv - 1) / p.ColThreshDiv;
        leftPctIdx = Math.Clamp(leftPctIdx, 0, leftExtents.Count - 1);
        rightPctIdx = Math.Clamp(rightPctIdx, 0, rightExtents.Count - 1);

        int bestColStart = leftExtents[leftPctIdx];
        int bestColEnd = rightExtents[rightPctIdx];

        Console.Error.WriteLine($"  edge-crop: horizontal: left={bestColStart} right={bestColEnd} ({bestColEnd - bestColStart + 1}px) samples={leftExtents.Count} pctL={leftPctIdx}/{leftExtents.Count} pctR={rightPctIdx}/{rightExtents.Count}");

        if (bestColEnd - bestColStart + 1 < 50)
        {
            Console.Error.WriteLine($"  edge-crop: horizontal extent too small");
            fullCapture.Dispose();
            return null;
        }

        // ── Phase 2: Per-column vertical extent ──
        int colBandHalf = (bestColEnd - bestColStart + 1) / 3;
        int colBandLeft = Math.Max(bestColStart, cursorX - colBandHalf);
        int colBandRight = Math.Min(bestColEnd, cursorX + colBandHalf);

        var topExtents = new List<int>();
        var bottomExtents = new List<int>();

        // Asymmetric gap: larger upward to bridge header decorations (~30-40px bright)
        int maxGapUp = maxGap * 3;

        for (int x = colBandLeft; x <= colBandRight; x++)
        {
            int ci = cursorY * stride + x * 4;
            int cBright = (px[ci] + px[ci + 1] + px[ci + 2]) / 3;
            if (cBright >= darkThresh) continue;

            int topEdge = cursorY;
            int gap = 0;
            for (int y = cursorY - 1; y >= 0; y--)
            {
                int i = y * stride + x * 4;
                int brightness = (px[i] + px[i + 1] + px[i + 2]) / 3;
                if (brightness < darkThresh) { topEdge = y; gap = 0; }
                else if (++gap > maxGapUp) break;
            }

            int bottomEdge = cursorY;
            gap = 0;
            for (int y = cursorY + 1; y < h; y++)
            {
                int i = y * stride + x * 4;
                int brightness = (px[i] + px[i + 1] + px[i + 2]) / 3;
                if (brightness < darkThresh) { bottomEdge = y; gap = 0; }
                else if (++gap > maxGap) break;
            }

            topExtents.Add(topEdge);
            bottomExtents.Add(bottomEdge);
        }

        if (topExtents.Count < 10)
        {
            Console.Error.WriteLine($"  edge-crop: too few dark columns ({topExtents.Count})");
            fullCapture.Dispose();
            return null;
        }

        topExtents.Sort();
        bottomExtents.Sort();

        int topPctIdx = topExtents.Count / p.RowThreshDiv;
        int botPctIdx = topExtents.Count * (p.ColThreshDiv - 1) / p.ColThreshDiv;
        topPctIdx = Math.Clamp(topPctIdx, 0, topExtents.Count - 1);
        botPctIdx = Math.Clamp(botPctIdx, 0, bottomExtents.Count - 1);

        int bestRowStart = topExtents[topPctIdx];
        int bestRowEnd = bottomExtents[botPctIdx];

        Console.Error.WriteLine($"  edge-crop: vertical: top={bestRowStart} bottom={bestRowEnd} ({bestRowEnd - bestRowStart + 1}px) samples={topExtents.Count}");

        if (bestRowEnd - bestRowStart + 1 < 50)
        {
            Console.Error.WriteLine($"  edge-crop: vertical extent too small");
            fullCapture.Dispose();
            return null;
        }

        int minX = bestColStart;
        int minY = bestRowStart;
        int maxX = bestColEnd;
        int maxY = bestRowEnd;

        int rw = maxX - minX + 1;
        int rh = maxY - minY + 1;

        Console.Error.WriteLine($"  edge-crop: result ({minX},{minY}) {rw}x{rh}");

        if (rw < 50 || rh < 50)
        {
            Console.Error.WriteLine($"  edge-crop: region too small ({rw}x{rh})");
            fullCapture.Dispose();
            return null;
        }

        var cropRect = new Rectangle(minX, minY, rw, rh);
        var cropped = fullCapture.Clone(cropRect, PixelFormat.Format32bppArgb);
        var region = new RegionRect { X = minX, Y = minY, Width = rw, Height = rh };

        return (cropped, fullCapture, region);
    }
}