UpScaler IP Core

Image scaling is a challenging task, especially when increasing resolution. The quality of the scaled image primarily depends on the interpolation method used to synthesize each pixel. The Lanczos method, and to a slightly lesser extent bicubic interpolation, are widely recognized as high-quality interpolation techniques. However, the complexity and specific requirements of these algorithms make them difficult to implement using FPGA technology.

The UpScaler IP Core utilizes an original algorithm that delivers quality comparable to bicubic interpolation while maintaining complexity close to that of linear interpolation.

The UpScaler IP Core is designed for use in video display systems such as video walls, displays, and projectors. It can also be applied in transcoding systems for preparing media content.

• The image quality is better than bilinear interpolation by 6-8 dB (according to the PSNR metric) and 1-1.5 dB better than the bicubic interpolation method.
• Supports RGB and YUV color spaces with an arbitrary bit depth for color components (the bit depth is set by the parameters of synthesis).
• Output resolution both horizontally and vertically is not limited.
• The maximum input horizontal resolution is determined by the volume of the internal buffer and set by the parameters of synthesis.
• The maximum input vertical resolution is not limited, though it will affect performance.
• Scaling coefficients are arbitrary, can be changed dynamically, and are independent horizontally and vertically. The bit depth of the coefficients is set by the parameters of synthesis.
• When specifying scaling coefficients, there is no need to load parameter tables.