Manually Doing Pixel Perfect Sprite Collisions (Examples & Video)
In this example we look over three methods for doing Pixel Perfect sprite collisions manually in program code, rather than using the built in sprite collision method (see SpriteCollisionMode) .
Sprite Collision Method 1: Raw Image (point by point) - The first example computes the common area the two sprites share and then reads the source pixels from the source sprite, checks for a mask colour then does the same for the Destination sprite. If the two pixels are Not mask colour, then we have a collision. This method works but there's a lot of surface overhead constantly pulling individual pixels from the surfaces via Point()
Sprite Collision Method 2: Byte Mask (Preprocess images into a 8bit Collision Mask) - The second method actually works almost the same as the first, except this time we create a collision mask of each image. The mask is an 8 bit version of the image, where each byte in the mask is either 0 for transparent or 1 for solid. The sprite collision routine then uses the same logic as method #1 but directly reads the mask bytes rather than reading pixels. This means it doesn't matter what type of image format (video/ FX/ AFX) the cost of reading the mask is the same regardless.
Sprite Collision Method #3: Span Buffer (Preprocess the images into a span list of hard pixels only) - The last method we look at is an extension to the mask concept, expect rather than store pixel data, we scan the image and make list of hard spans (none transparent pixels) for each row. Each span is just two values, the starting coordinate and length of the span. The data structure contains a table at the top of offsets into the span list, so we can look up each row as need be. Rows with no solid pixels, don't create any data, so we're only every comparing solid pixels to solid pixels. The regions might not overlap, but for most sprites it all boils down to only a couple of compares per row..
PlayBASIC has a number of sprite collision methods built in, ranging from simple bounding box intersections, rotated bounding box, Vector Shape ( Polygon Collision supporting Convex / Concave & complex polygons), Various sliding methods through to Pixel Perfect collision mode.
Video
Sample Code
This is the pixel perfect sprite collision function from the first demo, it basically just computes if the two sprites share a common area and then runs through the pixel data point() by point() exiting when it finds two pixels that aren't mask colour
In this example we look over three methods for doing Pixel Perfect sprite collisions manually in program code, rather than using the built in sprite collision method (see SpriteCollisionMode) .
Sprite Collision Method 1: Raw Image (point by point) - The first example computes the common area the two sprites share and then reads the source pixels from the source sprite, checks for a mask colour then does the same for the Destination sprite. If the two pixels are Not mask colour, then we have a collision. This method works but there's a lot of surface overhead constantly pulling individual pixels from the surfaces via Point()
Sprite Collision Method 2: Byte Mask (Preprocess images into a 8bit Collision Mask) - The second method actually works almost the same as the first, except this time we create a collision mask of each image. The mask is an 8 bit version of the image, where each byte in the mask is either 0 for transparent or 1 for solid. The sprite collision routine then uses the same logic as method #1 but directly reads the mask bytes rather than reading pixels. This means it doesn't matter what type of image format (video/ FX/ AFX) the cost of reading the mask is the same regardless.
Sprite Collision Method #3: Span Buffer (Preprocess the images into a span list of hard pixels only) - The last method we look at is an extension to the mask concept, expect rather than store pixel data, we scan the image and make list of hard spans (none transparent pixels) for each row. Each span is just two values, the starting coordinate and length of the span. The data structure contains a table at the top of offsets into the span list, so we can look up each row as need be. Rows with no solid pixels, don't create any data, so we're only every comparing solid pixels to solid pixels. The regions might not overlap, but for most sprites it all boils down to only a couple of compares per row..
PlayBASIC has a number of sprite collision methods built in, ranging from simple bounding box intersections, rotated bounding box, Vector Shape ( Polygon Collision supporting Convex / Concave & complex polygons), Various sliding methods through to Pixel Perfect collision mode.
Video
Sample Code
This is the pixel perfect sprite collision function from the first demo, it basically just computes if the two sprites share a common area and then runs through the pixel data point() by point() exiting when it finds two pixels that aren't mask colour
PlayBASIC Code:
; *=---------------------------------------------------------------------=* ; *=---------------------------------------------------------------------=* ; *=---------------------------------------------------------------------=* ;>> Sprite Hit Pixels << ; *=---------------------------------------------------------------------=* ; *=---------------------------------------------------------------------=* ; *=---------------------------------------------------------------------=* ; This function check if two spites overlap at pixel level. ; ; ; *=---------------------------------------------------------------------=* constant SpriteHitPixels_DEBUG = false function SpriteHitPixels(SrcSprite,DestSprite) SrcX1=GetSpriteX(SrcSprite)+GetSpriteHandleX(SrcSprite) SrcY1=GetSpriteY(SrcSprite)+GetSpriteHandleY(SrcSprite) SrcWidth=GetSpriteWidth(SrcSprite) SrcHeight=GetSpriteHeight(SrcSprite) DestX1=GetSpriteX(DestSprite)+GetSpriteHandleX(DestSprite) DestY1=GetSpriteY(DestSprite)+GetSpriteHandleY(DestSprite) DestWidth=GetSpriteWidth(DestSprite) DestHeight=GetSpriteHeight(DestSprite) // Manually check if the two sprites at least share the same area ; if (SrcX1+SrcWidth) > DestX1 if SrcX1<(DestX1+DestWidth) if (SrcY1+SrcHeight) > DestY1 if SrcY1<(DestY1+DestHeight) //Status=true // Scan through smallest region of the two OldSurface=GetSurface() SrcImage=GetSpriteImage(SrcSprite) SrcImageMaskColour =GetImageMaskColour(SrcImage) DestImage=GetSpriteImage(DestSprite) DestImageMaskColour =GetImageMaskColour(DestImage) // Compute the rect the two images share ClipX1=MaxVal(SrcX1,DestX1) ClipX2=MinVal(SrcX1+SrcWidth,DestX1+DestWidth) ClipY1=MaxVal(SrcY1,DestY1) ClipY2=MinVal(SrcY1+SrcHeight,DestY1+DestHeight) #IF SpriteHitPixels_DEBUG=true Boxc DestX1,DestY1,DestX1+DestWidth,DestY1+DestHeight,False, Rgb(255,0,0) Box ClipX1,ClipY1,ClipX2,ClipY2,false #ENDIF // Translate the world space cords to local // image space cordinates OffsetX =SrcX1 -DestX1 OffsetY =SrcY1 -DestY1 ClipX1 -=SrcX1 ClipX2 -=SrcX1 ClipY1 -=SrcY1 ClipY2 -=SrcY1 OldSurface=GetSurface() // Brute force scan through images for ScanLPY = ClipY1 to ClipY2 // Check if this row overlap rendertoimage SrcImage for ScanLPX = ClipX1 to ClipX2 ThisPixel1=Point(ScanLPX,ScanLPY) if ThisPixel1!=SrcImageMaskColour rendertoimage DestImage ThisPixel2=Point(OffsetX+ScanLPX,OffsetY+ScanLPY) if ThisPixel2!=DestImageMaskColour STATUS=TRUE #IF SpriteHitPixels_DEBUG=false eXITfor ScanLpY #endif #IF SpriteHitPixels_DEBUG=true rendertoimage 0 dotc OffsetX+ScanLPX,OffsetY+ScanLPY,$00ff00;ThisPixel2 #endif endif rendertoimage SrcImage endif next next rendertoimage OldSurface endif endif endif endif EndFunction Status
