Struct TransformOperation 

pub struct TransformOperation;

Transform operations for frequency domain processing

Implementations

impl TransformOperation

pub fn dct_2d( device: &GpuDevice, input: &[f32], output: &mut [f32], width: u32, height: u32, ) -> Result<()>

Compute 2D DCT (Discrete Cosine Transform)

Computes the forward DCT on 8x8 blocks. Input dimensions must be multiples of 8.

Arguments

• device - GPU device
• input - Input data (f32 values)
• output - Output DCT coefficients
• width - Data width (must be multiple of 8)
• height - Data height (must be multiple of 8)

Errors

Returns an error if dimensions are invalid or if the GPU operation fails.

pub fn idct_2d( device: &GpuDevice, input: &[f32], output: &mut [f32], width: u32, height: u32, ) -> Result<()>

Compute 2D IDCT (Inverse Discrete Cosine Transform)

Computes the inverse DCT on 8x8 blocks. Input dimensions must be multiples of 8.

Arguments

• device - GPU device
• input - Input DCT coefficients
• output - Output reconstructed data
• width - Data width (must be multiple of 8)
• height - Data height (must be multiple of 8)

Errors

Returns an error if dimensions are invalid or if the GPU operation fails.

pub fn dct_2d_general( device: &GpuDevice, input: &[f32], output: &mut [f32], width: u32, height: u32, ) -> Result<()>

Compute general 2D DCT using row-column decomposition

This method works for any dimensions, not just multiples of 8.

Arguments

• device - GPU device
• input - Input data (f32 values)
• output - Output DCT coefficients
• width - Data width
• height - Data height

Errors

Returns an error if dimensions are invalid or if the GPU operation fails.

impl TransformOperation

pub fn rotate90(data: &[u8], width: u32, height: u32, channels: u32) -> Vec<u8>

Rotate an interleaved pixel image 90° clockwise.

Output dimensions are swapped: out_width = height, out_height = width.

Pixel mapping: output(x, y) = input(y, width_out - 1 - x) where width_out = height.

Arguments

• data – packed pixel buffer (interleaved, channels bytes per pixel)
• width – source image width in pixels
• height – source image height in pixels
• channels – bytes per pixel (e.g. 3 for RGB, 4 for RGBA)

Panics

Panics in debug mode if data.len() != width * height * channels.

pub fn rotate180(data: &[u8], width: u32, height: u32, channels: u32) -> Vec<u8>

Rotate an interleaved pixel image 180°.

Output dimensions are the same as input.

Pixel mapping: output(x, y) = input(width-1-x, height-1-y).

pub fn rotate270(data: &[u8], width: u32, height: u32, channels: u32) -> Vec<u8>

Rotate an interleaved pixel image 270° clockwise (= 90° counter-clockwise).

Output dimensions are swapped: out_width = height, out_height = width.

Pixel mapping: output(x, y) = input(height-1-y, x).

pub fn flip_horizontal( data: &[u8], width: u32, height: u32, channels: u32, ) -> Vec<u8>

Flip an interleaved pixel image horizontally (mirror left-right).

Output dimensions are the same as input.

Pixel mapping: output(x, y) = input(width-1-x, y).

pub fn flip_vertical( data: &[u8], width: u32, height: u32, channels: u32, ) -> Vec<u8>

Flip an interleaved pixel image vertically (mirror top-bottom).

Output dimensions are the same as input.

Pixel mapping: output(x, y) = input(x, height-1-y).

pub fn transpose(data: &[u8], width: u32, height: u32, channels: u32) -> Vec<u8>

Transpose an interleaved pixel image (swap x and y axes).

Output dimensions are swapped: out_width = height, out_height = width.

Pixel mapping: output(x, y) = input(y, x).