Scalable Ultra-High Throughput Image Scaler

Overview

The UHT-SCALER core provides an image and video processing block which converts input images of one size and sampling format to output images of a different size and/or sampling format. It supports scaling of 4:4:4, 4:2:2, 4:2:0 and 4:0:0 (grayscale) image streams with 8- up to 16-bit color depth, using the BILINEAR, BICUBIC, LANCZOS and EXPFILTER scaling methods. The UHT-SCALER is a standalone and high-performance scaler core, designed for enabling high-quality and ultra-high throughput performance, even in low-end target silicon technologies. The UHT-SCALER is available for ASIC or AMD-Xilinx, Efinix, Intel, Lattice and Microchip FPGA and SoC based designs.

The core accepts the input image data in planar or interleaved scan order format and outputs the scaled image data in the same planar or interleaved format as the input.

The UHT-SCALER is very easy to use and integrate in a system. It requires minimal host intervention as it only needs to be programmed once per video sequence. Once programmed, it can scale an arbitrary number of video frames without the need of any further intervention or assistance by the host system CPU.

Key Features

Standalone Scaling Operation
- Programmable image dimensions from 8 x 8 up to 64k x 64k
- Supports YcbCr/RGB 4:4:4, YCbCr 4:2:2, YCbCr 4:2:0 and single-color 4:0:0 video formats
- 8 up to 16 bits per sample depth encoding
- Support for bilinear, bicubic, lanczos and expfilter scaling algorithms
- CPU-less, complete and standalone operation.
Advanced Scaling Implementation
- Ultra-high throughput in low-end silicon
- Ultra-low latency performance
- On-chip memory implementation
Easy Implementation and Verification
- Extensive documentation
- Bit Accurate Model (BAM) with optional Test Vector generation functionality
- Self-checking testbench environment
- Sample BAM scripts
- Synthesis scripts
- Simulation scripts
- Place & Route scripts for FPGAs
Trouble-Free Technology Map and Implementation
- Self-contained RTL design
- No internal tri-states
- Strictly positive edge triggered design
- D-type only Flip-Flops
- Fully synchronous operation per clock domain¹
- Safe CDC transfers between clock domains²
- No need for special timing constraints
  - No false or multi-cycle paths within the same clock domain
  - No CDC transfers that need to be specially constrained
  - No other specially constrained timing paths

Block Diagram

Deliverables

Clear-text RTL sources for ASIC designs, or pre-synthesized and verified Netlist for FPGA and SoC devices
Release Notes, Design Specification and Integration Manual documents
Bit Accurate Model (BAM) and test vector generation binaries, including sample scripts
Pre-compiled RTL simulation model and gate-level simulation netlist for the FPGA Netlist license
Self-checking testbench environment sources, including sample BAM generated test cases
Simulation and sample Synthesis (for ASICs) or Place & Route (for FPGAs) scripts

Technical Specifications

Request Info