The inflated bytes are not pixels yet - they are filtered scanlines, each row prefixed by one filter-type byte. Today you work out the row layout and the bytes-per-pixel value every filter needs.
Compute the bytes per pixel and the filtered stride of a scanline from the image header.
The bytes coming out of Inflate are organized as scanlines, one per image row, and each row begins with a single filter-type byte that says how that row was transformed before compression. Filtering makes pixel data more compressible by storing differences from neighbors rather than raw values, and it is undone one row at a time. Before you can undo it you need two numbers: the bytes per pixel (bpp) and the stride, the total byte length of a filtered row.
Bytes per pixel is the channel count times the bit depth over 8, rounded up and floored at 1 - so 8-bit RGBA is 4, and a 1-bit grayscale pixel still counts as 1 byte for the filter’s stepping purposes. The stride is 1 + row-data-bytes: the leading filter byte plus the packed samples. These values drive every filter that follows, because Sub and Paeth reach back exactly bpp bytes to find the pixel to the left. Get them right and the reconstruction filters have a solid frame to work in.
// channels: gray=1, rgb=3, palette=1, gray+alpha=2, rgba=4.// bpp = ceil(channels * bitDepth / 8), min 1.// filtered stride = 1 + ceil(width * channels * bitDepth / 8)func bytesPerPixel(colorType, bitDepth int) int { }