linesizes

Linesizes, often referred to as strides or pitches, describe the number of bytes between the start of one row of image data and the start of the next row in memory. They are used to navigate two-dimensional image buffers stored in a one-dimensional memory layout. The linesize value is typically provided by the graphics or video API and may differ between formats and planes.

In packed formats, where pixels are stored consecutively (for example, RGB24), the linesize usually equals the row width in bytes, possibly with padding for alignment. In planar formats, such as many YUV schemes, there is a separate linesize for each plane. For instance, in a common YUV 4:2:0 layout, the luma plane may have linesize equal to the width in bytes, while the chroma planes have their own linesizes, often smaller due to subsampling.

Linesizes allow correct row-by-row access even when memory alignment introduces padding. A common usage pattern is to locate a pixel at coordinates (x,y) in a given plane by offsetting the base pointer by y times linesize plus x times bytes per pixel for packed data, or by y times the plane’s linesize for planar data.

Notes and nuances: linesize is not necessarily equal to image width times bytes per pixel; it may include padding to satisfy alignment constraints (e.g., to a multiple of 4, 8, or 16). Some APIs also support negative linesizes to indicate certain storage orders (such as bottom-up). In practice, linesizes are API-provided attributes that should be relied upon rather than inferred from image dimensions.

Common contexts include video decoding and encoding libraries (for example, FFmpeg) and graphics APIs, where correct linesize handling is essential for accurate image processing.