Gmarker689
f3dfcd3b0b
- 新增 RemoveWhiteBg：纯白色背景(#FFFFFF)像素半透明化 - 新增 fixedGridSplit：GridRows×GridCols 固定网格拆分 - 新增 CenterAlign：帧内容居中对齐，确保人物不跳帧 - Options 新增 WhiteBg/GridRows/GridCols/GridPadding/CenterAlign - 默认模式从绿幕切换为白底 - 新增 tools/test_prompt_to_gif.go 端到端测试脚本 - .gitignore 忽略 test_output 目录
482 lines
12 KiB
Go
Executable File
482 lines
12 KiB
Go
Executable File
// Package splitsprite provides PNG sprite sheet splitting utilities.
|
||
//
|
||
// Pipeline: white/green background removal → projection-based gap detection →
|
||
// split into tiles → filter out low-fill tiles → trim transparent edges.
|
||
package splitsprite
|
||
|
||
import (
|
||
"fmt"
|
||
"image"
|
||
"image/color"
|
||
"image/draw"
|
||
"math"
|
||
"sort"
|
||
)
|
||
|
||
// Options configures the sprite sheet splitting pipeline.
|
||
type Options struct {
|
||
// WhiteBg enables white background removal.
|
||
WhiteBg bool
|
||
// WhiteThreshold is the max distance from pure white (0–255, default 40).
|
||
WhiteThreshold uint8
|
||
|
||
// GreenScreen enables green background removal.
|
||
GreenScreen bool
|
||
// GreenTolerance controls how aggressively green pixels are removed (0–1, default 0.2).
|
||
GreenTolerance float64
|
||
|
||
// GridRows / GridCols enable fixed-grid splitting (overrides projection detection).
|
||
// When >0, the image is divided equally into Rows×Cols cells.
|
||
GridRows int
|
||
GridCols int
|
||
// GridPadding is the gap between cells in pixels (default 2).
|
||
GridPadding int
|
||
|
||
// GapThreshold is the max fraction of non-transparent pixels a row/column
|
||
// can have to be considered a gap (0–1, default 0.03).
|
||
GapThreshold float64
|
||
// MinGapWidth is the minimum width in pixels a gap must have.
|
||
MinGapWidth int
|
||
|
||
// MinFillRatio is the minimum fraction of non-transparent pixels a tile
|
||
// must have to be kept (0–1, default 0.3).
|
||
MinFillRatio float64
|
||
|
||
// Trim removes transparent borders from output tiles.
|
||
Trim bool
|
||
// CenterAlign centers content across all frames so characters stay in place.
|
||
// All output frames get the same dimensions with content centered.
|
||
CenterAlign bool
|
||
// OutW / OutH specify the output tile size (0 = keep original).
|
||
OutW, OutH int
|
||
}
|
||
|
||
// DefaultOptions returns sensible default splitting options (white background mode).
|
||
func DefaultOptions() *Options {
|
||
return &Options{
|
||
WhiteBg: true,
|
||
WhiteThreshold: 40,
|
||
GapThreshold: 0.03,
|
||
MinGapWidth: 2,
|
||
MinFillRatio: 0.3,
|
||
Trim: true,
|
||
}
|
||
}
|
||
|
||
// DefaultGreenOptions returns options tuned for green screen sprite sheets.
|
||
func DefaultGreenOptions() *Options {
|
||
return &Options{
|
||
GreenScreen: true,
|
||
GreenTolerance: 0.2,
|
||
GapThreshold: 0.03,
|
||
MinGapWidth: 2,
|
||
MinFillRatio: 0.3,
|
||
Trim: true,
|
||
}
|
||
}
|
||
|
||
// Process splits a sprite sheet image into individual cleaned tile images.
|
||
// It runs the full pipeline: background removal → split → trim → resize.
|
||
// When GridRows/GridCols > 0, fixed-grid splitting is used instead of projection detection.
|
||
func Process(img image.Image, opts *Options) ([]image.Image, error) {
|
||
if opts == nil {
|
||
opts = DefaultOptions()
|
||
}
|
||
|
||
src := toRGBA(img)
|
||
|
||
if opts.WhiteBg {
|
||
src = removeWhiteBg(src, opts.WhiteThreshold)
|
||
} else if opts.GreenScreen {
|
||
src = removeGreenScreen(src, opts.GreenTolerance)
|
||
}
|
||
|
||
var tiles []tile
|
||
if opts.GridRows > 0 && opts.GridCols > 0 {
|
||
tiles = fixedGridSplit(src, opts.GridRows, opts.GridCols, opts.GridPadding, opts.MinFillRatio)
|
||
} else {
|
||
tiles = projectionSplit(src, opts.GapThreshold, opts.MinGapWidth, opts.MinFillRatio)
|
||
}
|
||
if len(tiles) == 0 {
|
||
return nil, fmt.Errorf("no tiles detected — try lowering GapThreshold or setting GridRows/GridCols")
|
||
}
|
||
|
||
results := make([]image.Image, len(tiles))
|
||
for i, t := range tiles {
|
||
sub := image.NewRGBA(image.Rect(0, 0, t.w, t.h))
|
||
draw.Draw(sub, sub.Bounds(), src, image.Point{t.x, t.y}, draw.Src)
|
||
|
||
if opts.Trim {
|
||
sub = trimAlpha(sub)
|
||
}
|
||
if opts.OutW > 0 && opts.OutH > 0 {
|
||
sub = resize(sub, opts.OutW, opts.OutH)
|
||
}
|
||
results[i] = sub
|
||
}
|
||
|
||
if opts.CenterAlign && len(results) > 1 {
|
||
results = alignCenter(results)
|
||
}
|
||
|
||
return results, nil
|
||
}
|
||
|
||
// RemoveWhiteBg removes near-white background pixels, making them transparent.
|
||
func RemoveWhiteBg(img image.Image, threshold uint8) image.Image {
|
||
return removeWhiteBg(toRGBA(img), threshold)
|
||
}
|
||
|
||
// CenterFrames centers the content of each frame within a uniform canvas so
|
||
// characters stay in place across frames.
|
||
func CenterFrames(frames []image.Image) []image.Image {
|
||
return alignCenter(frames)
|
||
}
|
||
|
||
// alignCenter finds the content bounding box per frame, computes the max
|
||
// dimensions, then pads each frame so content is centered uniformly.
|
||
func alignCenter(frames []image.Image) []image.Image {
|
||
type contentBox struct {
|
||
minX, minY, maxX, maxY int
|
||
}
|
||
boxes := make([]contentBox, len(frames))
|
||
maxW, maxH := 0, 0
|
||
|
||
for i, f := range frames {
|
||
b := f.Bounds()
|
||
minX, minY := b.Max.X, b.Max.Y
|
||
maxX, maxY := b.Min.X, b.Min.Y
|
||
hasContent := false
|
||
for y := b.Min.Y; y < b.Max.Y; y++ {
|
||
for x := b.Min.X; x < b.Max.X; x++ {
|
||
_, _, _, a := f.At(x, y).RGBA()
|
||
if a > 0 {
|
||
hasContent = true
|
||
if x < minX {
|
||
minX = x
|
||
}
|
||
if x > maxX {
|
||
maxX = x
|
||
}
|
||
if y < minY {
|
||
minY = y
|
||
}
|
||
if y > maxY {
|
||
maxY = y
|
||
}
|
||
}
|
||
}
|
||
}
|
||
if !hasContent {
|
||
boxes[i] = contentBox{0, 0, b.Dx(), b.Dy()}
|
||
} else {
|
||
boxes[i] = contentBox{minX, minY, maxX, maxY}
|
||
}
|
||
w := boxes[i].maxX - boxes[i].minX + 1
|
||
h := boxes[i].maxY - boxes[i].minY + 1
|
||
if w > maxW {
|
||
maxW = w
|
||
}
|
||
if h > maxH {
|
||
maxH = h
|
||
}
|
||
}
|
||
|
||
// Pad by 10% to avoid edge cropping
|
||
maxW = maxW * 11 / 10
|
||
maxH = maxH * 11 / 10
|
||
|
||
out := make([]image.Image, len(frames))
|
||
for i, f := range frames {
|
||
cb := boxes[i]
|
||
cw := cb.maxX - cb.minX + 1
|
||
ch := cb.maxY - cb.minY + 1
|
||
ox := (maxW - cw) / 2
|
||
oy := (maxH - ch) / 2
|
||
|
||
canvas := image.NewRGBA(image.Rect(0, 0, maxW, maxH))
|
||
draw.Draw(canvas,
|
||
image.Rect(ox, oy, ox+cw, oy+ch),
|
||
f,
|
||
image.Point{cb.minX, cb.minY},
|
||
draw.Src,
|
||
)
|
||
out[i] = canvas
|
||
}
|
||
return out
|
||
}
|
||
|
||
// RemoveGreenScreen removes green-dominant background pixels, making them transparent.
|
||
func RemoveGreenScreen(img image.Image, tol float64) image.Image {
|
||
return removeGreenScreen(toRGBA(img), tol)
|
||
}
|
||
|
||
// TrimAlpha removes fully transparent borders from an image.
|
||
func TrimAlpha(img image.Image) image.Image {
|
||
return trimAlpha(toRGBA(img))
|
||
}
|
||
|
||
// Resize resizes an image using nearest-neighbor interpolation.
|
||
func Resize(img image.Image, w, h int) image.Image {
|
||
return resize(toRGBA(img), w, h)
|
||
}
|
||
|
||
// ============================
|
||
// internal
|
||
// ============================
|
||
|
||
type tile struct {
|
||
x, y, w, h int
|
||
}
|
||
|
||
// removeWhiteBg removes pixels close to pure white (R,G,B all above threshold).
|
||
func removeWhiteBg(rgba *image.RGBA, threshold uint8) *image.RGBA {
|
||
if threshold == 0 {
|
||
threshold = 40
|
||
}
|
||
b := rgba.Bounds()
|
||
dst := image.NewRGBA(b)
|
||
draw.Draw(dst, b, rgba, b.Min, draw.Src)
|
||
|
||
for y := b.Min.Y; y < b.Max.Y; y++ {
|
||
for x := b.Min.X; x < b.Max.X; x++ {
|
||
r, g, bl, a := rgba.At(x, y).RGBA()
|
||
if a == 0 {
|
||
continue
|
||
}
|
||
r8, g8, b8 := uint8(r>>8), uint8(g>>8), uint8(bl>>8)
|
||
// Pixel is "white" when all channels are near 255
|
||
if int(255-r8) < int(threshold) && int(255-g8) < int(threshold) && int(255-b8) < int(threshold) {
|
||
// Calculate alpha: closer to white = more transparent
|
||
dist := max(int(255-r8), max(int(255-g8), int(255-b8)))
|
||
alpha := float64(dist) / float64(threshold)
|
||
dst.SetRGBA(x, y, color.RGBA{R: r8, G: g8, B: b8, A: uint8(alpha * 255)})
|
||
}
|
||
}
|
||
}
|
||
return dst
|
||
}
|
||
|
||
func removeGreenScreen(rgba *image.RGBA, tol float64) *image.RGBA {
|
||
b := rgba.Bounds()
|
||
dst := image.NewRGBA(b)
|
||
absTol := tol * 255
|
||
|
||
for y := b.Min.Y; y < b.Max.Y; y++ {
|
||
for x := b.Min.X; x < b.Max.X; x++ {
|
||
r16, g16, bl16, a16 := rgba.At(x, y).RGBA()
|
||
if a16 == 0 {
|
||
continue
|
||
}
|
||
r, g, bl := float64(r16>>8), float64(g16>>8), float64(bl16>>8)
|
||
gDominance := g - (r+bl)/2
|
||
if gDominance > absTol {
|
||
alpha := 1.0 - math.Min(gDominance/(absTol*2), 1.0)
|
||
dst.SetRGBA(x, y, color.RGBA{
|
||
R: uint8(r), G: uint8(g), B: uint8(bl),
|
||
A: uint8(alpha * 255),
|
||
})
|
||
} else {
|
||
dst.Set(x, y, rgba.At(x, y))
|
||
}
|
||
}
|
||
}
|
||
return dst
|
||
}
|
||
|
||
// fixedGridSplit divides the image into Rows×Cols equally-sized cells,
|
||
// accounting for a fixed padding between cells.
|
||
func fixedGridSplit(rgba *image.RGBA, rows, cols, padding int, minFill float64) []tile {
|
||
b := rgba.Bounds()
|
||
W, H := b.Dx(), b.Dy()
|
||
|
||
if padding < 0 {
|
||
padding = 0
|
||
}
|
||
totalPadW := padding * (cols + 1)
|
||
totalPadH := padding * (rows + 1)
|
||
cellW := (W - totalPadW) / cols
|
||
cellH := (H - totalPadH) / rows
|
||
if cellW <= 0 || cellH <= 0 {
|
||
return nil
|
||
}
|
||
|
||
var tiles []tile
|
||
for r := 0; r < rows; r++ {
|
||
for c := 0; c < cols; c++ {
|
||
x := padding + c*(cellW+padding)
|
||
y := padding + r*(cellH+padding)
|
||
if tileFillRatio(rgba, x, y, cellW, cellH) >= minFill {
|
||
tiles = append(tiles, tile{x: x, y: y, w: cellW, h: cellH})
|
||
}
|
||
}
|
||
}
|
||
return tiles
|
||
}
|
||
|
||
func projectionSplit(rgba *image.RGBA, gapThreshold float64, minGap int, minFill float64) []tile {
|
||
bounds := rgba.Bounds()
|
||
W, H := bounds.Dx(), bounds.Dy()
|
||
|
||
rowRatio := make([]float64, H)
|
||
colRatio := make([]float64, W)
|
||
for y := 0; y < H; y++ {
|
||
n := 0
|
||
for x := 0; x < W; x++ {
|
||
if alphaAt(rgba, x, y) > 0 {
|
||
n++
|
||
}
|
||
}
|
||
rowRatio[y] = float64(n) / float64(W)
|
||
}
|
||
for x := 0; x < W; x++ {
|
||
n := 0
|
||
for y := 0; y < H; y++ {
|
||
if alphaAt(rgba, x, y) > 0 {
|
||
n++
|
||
}
|
||
}
|
||
colRatio[x] = float64(n) / float64(H)
|
||
}
|
||
|
||
rowCuts := findCuts(rowRatio, gapThreshold, minGap)
|
||
colCuts := findCuts(colRatio, gapThreshold, minGap)
|
||
|
||
if len(rowCuts) < 2 || len(colCuts) < 2 {
|
||
return nil
|
||
}
|
||
|
||
var tiles []tile
|
||
for ri := 0; ri < len(rowCuts)-1; ri++ {
|
||
for ci := 0; ci < len(colCuts)-1; ci++ {
|
||
x := colCuts[ci]
|
||
y := rowCuts[ri]
|
||
w := colCuts[ci+1] - x
|
||
h := rowCuts[ri+1] - y
|
||
if tileFillRatio(rgba, x, y, w, h) >= minFill {
|
||
tiles = append(tiles, tile{x: x, y: y, w: w, h: h})
|
||
}
|
||
}
|
||
}
|
||
return tiles
|
||
}
|
||
|
||
func tileFillRatio(rgba *image.RGBA, x0, y0, w, h int) float64 {
|
||
total := w * h
|
||
if total == 0 {
|
||
return 0
|
||
}
|
||
n := 0
|
||
for y := y0; y < y0+h; y++ {
|
||
for x := x0; x < x0+w; x++ {
|
||
if alphaAt(rgba, x, y) > 0 {
|
||
n++
|
||
}
|
||
}
|
||
}
|
||
return float64(n) / float64(total)
|
||
}
|
||
|
||
func findCuts(ratios []float64, threshold float64, minGap int) []int {
|
||
n := len(ratios)
|
||
isGap := make([]bool, n)
|
||
for i, r := range ratios {
|
||
isGap[i] = r < threshold
|
||
}
|
||
|
||
type segment struct{ start, end int }
|
||
var gaps []segment
|
||
i := 0
|
||
for i < n {
|
||
if isGap[i] {
|
||
start := i
|
||
for i < n && isGap[i] {
|
||
i++
|
||
}
|
||
if i-start >= minGap {
|
||
gaps = append(gaps, segment{start, i})
|
||
}
|
||
} else {
|
||
i++
|
||
}
|
||
}
|
||
|
||
if len(gaps) == 0 {
|
||
return []int{0, n}
|
||
}
|
||
|
||
cuts := []int{0}
|
||
for _, seg := range gaps {
|
||
cuts = append(cuts, seg.start+(seg.end-seg.start)/2)
|
||
}
|
||
cuts = append(cuts, n)
|
||
sort.Ints(cuts)
|
||
|
||
dedup := cuts[:1]
|
||
for j := 1; j < len(cuts); j++ {
|
||
if cuts[j] != dedup[len(dedup)-1] {
|
||
dedup = append(dedup, cuts[j])
|
||
}
|
||
}
|
||
return dedup
|
||
}
|
||
|
||
func alphaAt(rgba *image.RGBA, x, y int) uint8 {
|
||
return rgba.Pix[rgba.PixOffset(x, y)+3]
|
||
}
|
||
|
||
func trimAlpha(rgba *image.RGBA) *image.RGBA {
|
||
b := rgba.Bounds()
|
||
minX, minY := b.Max.X, b.Max.Y
|
||
maxX, maxY := b.Min.X, b.Min.Y
|
||
for y := b.Min.Y; y < b.Max.Y; y++ {
|
||
for x := b.Min.X; x < b.Max.X; x++ {
|
||
if alphaAt(rgba, x-b.Min.X, y-b.Min.Y) > 0 {
|
||
if x < minX {
|
||
minX = x
|
||
}
|
||
if x > maxX {
|
||
maxX = x
|
||
}
|
||
if y < minY {
|
||
minY = y
|
||
}
|
||
if y > maxY {
|
||
maxY = y
|
||
}
|
||
}
|
||
}
|
||
}
|
||
w := maxX - minX + 1
|
||
h := maxY - minY + 1
|
||
if w <= 0 || h <= 0 {
|
||
return rgba
|
||
}
|
||
dst := image.NewRGBA(image.Rect(0, 0, w, h))
|
||
draw.Draw(dst, dst.Bounds(), rgba, image.Point{minX, minY}, draw.Src)
|
||
return dst
|
||
}
|
||
|
||
func resize(rgba *image.RGBA, w, h int) *image.RGBA {
|
||
dst := image.NewRGBA(image.Rect(0, 0, w, h))
|
||
sw, sh := rgba.Bounds().Dx(), rgba.Bounds().Dy()
|
||
for y := 0; y < h; y++ {
|
||
for x := 0; x < w; x++ {
|
||
sx := x * sw / w
|
||
sy := y * sh / h
|
||
dst.Set(x, y, rgba.At(sx, sy))
|
||
}
|
||
}
|
||
return dst
|
||
}
|
||
|
||
func toRGBA(src image.Image) *image.RGBA {
|
||
if rgba, ok := src.(*image.RGBA); ok {
|
||
return rgba
|
||
}
|
||
b := src.Bounds()
|
||
rgba := image.NewRGBA(b)
|
||
draw.Draw(rgba, b, src, b.Min, draw.Src)
|
||
return rgba
|
||
}
|