KOPTContext: add comic panel detection function/additional leptonica declarations (koreader#1159)

Used to find manga/comic panels. It uses connected component labeling,
for detection. The downside of this algorithm is that panels joined by
text, "word clouds" etc. are detected as one entity. Compeletely open
(not having borders) panels may sometimes fail to register.

Author: Galunid

ffi-cdecl/leptonica_cdecl.c

@@ -14,6 +14,9 @@ cdecl_type(l_float64)
 
 cdecl_struct(Box)
 cdecl_type(BOX)
+cdecl_struct(Sel)
+cdecl_type(Sel)
+
 cdecl_struct(Boxa)
 cdecl_type(BOXA)
 cdecl_struct(Numa)
@@ -35,6 +38,7 @@ cdecl_func(numaCreate)
 cdecl_func(numaCreateFromFArray)
 cdecl_func(boxaCombineOverlaps)
 cdecl_func(boxaClipToBox)
+cdecl_func(pixConnCompBB)
 cdecl_func(boxCopy)
 cdecl_func(boxClone)
 cdecl_func(boxOverlapRegion)
@@ -58,12 +62,25 @@ cdecl_func(pixGetWpl)
 cdecl_func(pixSetPixel)
 cdecl_func(pixGetData)
 cdecl_func(pixCreate)
+cdecl_func(pixClone)
 cdecl_func(pixConvertTo1)
 cdecl_func(pixThresholdToBinary)
+cdecl_func(pixConvertRGBToGrayFast)
 cdecl_func(pixConvertTo32)
 cdecl_func(pixDrawBoxaRandom)
 cdecl_func(pixMultiplyByColor)
 cdecl_func(pixBlendBackgroundToColor)
+cdecl_func(pixBlockconv)
+cdecl_func(pixRenderContours)
+cdecl_func(pixInvert)
+cdecl_func(pixClipRectangle)
+cdecl_func(pixOpen)
+cdecl_func(pixClose)
+cdecl_func(pixErode)
 cdecl_func(pixGetRegionsBinary)
 cdecl_func(pixSplitIntoBoxa)
 cdecl_func(pixReduceRankBinaryCascade)
+cdecl_func(selCreate)
+cdecl_func(selSetElement)
+cdecl_func(selPrintToString)
+cdecl_func(selDestroy)

ffi/koptcontext.lua

@@ -257,6 +257,66 @@ function KOPTContext_mt.__index:findPageBlocks()
     end
 end
 
+function KOPTContext_mt.__index:getPanelFromPage(pos)
+    local function isInRect(x, y, w, h, pos_x, pos_y)
+        return x < pos_x and y < pos_y and x + w > pos_x and y + h > pos_y
+    end
+
+    if self.src.data then
+        local pixs = k2pdfopt.bitmap2pix(self.src, 0, 0, self.src.width, self.src.height)
+        local pixg
+        if leptonica.pixGetDepth(pixs) == 32 then
+            pixg = leptonica.pixConvertRGBToGrayFast(pixs)
+        else
+            pixg = leptonica.pixClone(pixs)
+        end
+
+        -- leptonica's threshold gets pixels lighter than X, we want to get
+        -- pixels darker than X, to do that we invert the image, threshold it,
+        -- and invert the result back. Math: ~(~img < X) <=> img > X
+        local pix_inverted = leptonica.pixInvert(nil, pixg)
+        local pix_thresholded = leptonica.pixThresholdToBinary(pix_inverted, 50)
+        leptonica.pixInvert(pix_thresholded, pix_thresholded)
+
+        -- find connected components (in our case panels)
+        local bb = leptonica.pixConnCompBB(pix_thresholded, 8)
+
+        local img_w = leptonica.pixGetWidth(pixs)
+        local img_h = leptonica.pixGetHeight(pixs)
+        local res
+
+        for i = 0, leptonica.boxaGetCount(bb) - 1 do
+            local box = leptonica.boxaGetBox(bb, i, C.L_CLONE)
+            local pix_tmp = leptonica.pixClipRectangle(pixs, box, nil)
+            local w = leptonica.pixGetWidth(pix_tmp)
+            local h = leptonica.pixGetHeight(pix_tmp)
+            -- check if it's panel or part of the panel, if it's part of the panel skip
+            if w >= img_w / 8 and h >= img_h / 8 then
+                if isInRect(box.x, box.y, box.w, box.h, pos.x, pos.y) then
+                    res = {
+                        x = box.x,
+                        y = box.y,
+                        w = box.w,
+                        h = box.h,
+                    }
+                    leptonica.pixDestroy(ffi.new("PIX *[1]", pix_tmp))
+                    leptonica.boxDestroy(ffi.new("BOX *[1]", box))
+                    break -- we found panel, exit the loop and clean up memory
+                end
+            end
+            leptonica.pixDestroy(ffi.new("PIX *[1]", pix_tmp))
+            leptonica.boxDestroy(ffi.new("BOX *[1]", box))
+        end
+
+        -- free up memory
+        leptonica.boxaDestroy(ffi.new("BOXA *[1]", bb))
+        leptonica.pixDestroy(ffi.new("PIX *[1]", pixg))
+        leptonica.pixDestroy(ffi.new("PIX *[1]", pix_thresholded))
+        leptonica.pixDestroy(ffi.new("PIX *[1]", pixs))
+        return res
+    end
+end
+
 --[[
 -- get page block in location x, y both of which in range [0, 1] relative to page
 -- width and height respectively

ffi/leptonica_h.lua

@@ -59,13 +59,23 @@ struct Pix {
   l_uint32 *data;
 };
 typedef struct Pix PIX;
+struct Sel {
+  l_int32 sx;
+  l_int32 sy;
+  l_int32 cx;
+  l_int32 cy;
+  l_uint32 **data;
+  char *name;
+};
+typedef struct Sel SEL;
 BOX *boxCreate(l_int32, l_int32, l_int32, l_int32);
 BOXA *boxaCreate(l_int32);
 BOX *boxaGetBox(BOXA *, l_int32, l_int32);
 NUMA *numaCreate(l_int32);
 NUMA *numaCreateFromFArray(l_float32 *, l_int32, l_int32);
 BOXA *boxaCombineOverlaps(BOXA *);
 BOXA *boxaClipToBox(BOXA *, BOX *);
+BOXA *pixConnCompBB(PIX *, l_uint32);
 BOX *boxCopy(BOX *);
 BOX *boxClone(BOX *);
 BOX *boxOverlapRegion(BOX *, BOX *);
@@ -89,13 +99,26 @@ l_int32 pixGetWpl(PIX *);
 l_int32 pixSetPixel(PIX *, l_int32, l_int32, l_uint32);
 l_uint32 *pixGetData(PIX *);
 PIX *pixCreate(l_int32, l_int32, l_int32);
+PIX *pixClone(PIX *);
 PIX *pixConvertTo1(PIX *, l_int32);
 PIX *pixThresholdToBinary(PIX *, l_int32);
+PIX *pixConvertRGBToGrayFast(PIX *);
 PIX *pixConvertTo32(PIX *);
 PIX *pixDrawBoxaRandom(PIX *, BOXA *, l_int32);
 PIX *pixMultiplyByColor(PIX *, PIX *, BOX *, l_uint32);
 PIX *pixBlendBackgroundToColor(PIX *, PIX *, BOX *, l_uint32, l_float32, l_int32, l_int32);
+PIX *pixBlockconv(PIX *, l_uint32, l_uint32);
+PIX *pixRenderContours(PIX *, l_uint32, l_uint32, l_uint32);
+PIX *pixInvert(PIX *, PIX *);
+PIX *pixClipRectangle(PIX *, BOX *, BOX **);
+PIX *pixOpen(PIX *, PIX *, SEL *);
+PIX *pixClose(PIX *, PIX *, SEL *);
+PIX *pixErode(PIX *, PIX *, SEL *);
 l_int32 pixGetRegionsBinary(PIX *, PIX **, PIX **, PIX **, struct Pixa *);
 BOXA *pixSplitIntoBoxa(PIX *, l_int32, l_int32, l_int32, l_int32, l_int32, l_int32);
 PIX *pixReduceRankBinaryCascade(PIX *, l_int32, l_int32, l_int32, l_int32);
+SEL *selCreate(l_int32, l_int32, char *);
+l_int32 selSetElement(SEL *, l_int32, l_int32, l_int32);
+char *selPrintToString(SEL *);
+void selDestroy(SEL **);
 ]]