Assumes that an initial valid guess of change rectangle 'rect' is passed.

Get the canvas width, height and flags after validating that VP8X/VP8/VP8L chunk and canvas size are valid.

Size of an image.

Adjusts pixel values of image with given maximum error.

Check if we need to preserve the compressed alpha data, as it may not have been decoded yet.

Check if we need to preserve the compressed alpha data, as it may not have been decoded yet.

input pixels are uint8_t

Fetch '*width', '*height', '*has_alpha' and fill out 'headers' based on 'data'. All the output parameters may be NULL. If 'headers' is NULL only the m

Fetch '*width', '*height', '*has_alpha' and fill out 'headers' based on 'data'. All the output parameters may be NULL. If 'headers' is NULL only the m

Pop an interval in the manager.

Given a current orphan interval and its previous interval, before it was orphaned (which can be NULL), set it at the right place in the list of interv

Sort palette in increasing order and prepare an inverse mapping array.

Given a new cost interval defined by its start at position, its length value and distance_cost, add its contributions to the previous intervals and co

Given a new cost interval defined by its start at index, its last value and distance_cost, add its contributions to the previous intervals and costs.

Unpack the 8b input rgbrgbrgbrgb ... as contiguous registers: rrrr... rrrr... gggg... gggg... bbbb... bbbb.... Similar to PlanarTo24bHelper(), but in

Unpack the 8b input rgbrgbrgbrgb ... as contiguous registers: rrrr... rrrr... gggg... gggg... bbbb... bbbb.... Similar to PlanarTo24bHelper(), but in

Convert 16 packed ARGB 16b-values to r[], g[], b[]

Convert 16 packed ARGB 16b-values to r[], g[], b[]

Read from 'mem' and skip the read bytes.

'code_lengths' is pre-allocated temporary buffer, used for creating Huffman tree.

Outputs a pointer to 'prev_wpi->next_', where 'prev_wpi' is the pointer to the image at position (nth - 1). Returns true if nth image was found.

Extracts image & alpha data from the given bitstream and then sets wpi.alpha_ and wpi.img_ appropriately.

If not at EOS, reload up to VP8L_LBITS byte-by-byte

If not at EOS, reload up to VP8L_LBITS byte-by-byte

Snap rectangle to even offsets (and adjust dimensions if needed).

Delta palettization functions.

Store image bearing chunks to 'frame'. 'min_size' is an optional size requirement, it may be zero.

'huff_tree' and 'tokens' are pre-alloacted buffers.

'huff_tree' and 'tokens' are pre-alloacted buffers.

Assemble a single image WebP bitstream from 'wpi'.

Hadamard transform Returns the weighted sum of the absolute value of transformed coefficients. w[] contains a row-major 4 by 4 symmetric matrix.

TrueMotion (4x4 + 8x8)