Build Image Processing Pipeline using FPGA IPs

In addition to high speed image capture using our proprietary CameraLink IP our Falcon products also have the capability to process images using our proprietary FPGA IPs. These IPs can be easily integrated from software using a simple Python API call.

Our FPGA IPs can be either used for pre-processing to reduce bandwidth or for building a complete image processing pipeline for hard real-time applications.

Image Processing IP – Binarization & Quantization

Binarization & Quantization IP is used in the image processing pipeline to binarize the image and quantize to lesser no. of bits hence reducing the bandwidth required to process huge volume of images.

Image Processing IP – Color Space Convertor

Color Space converter IP is used in an image processing pipeline to convert color space of images between RGB, HSV and YUV depending on the requirement of the application.

Image Processing IP – Contour Tracing

Contour Tracing IP is a segmentation technique used to identify the boundary region of a particular region. The IP takes binarized images as input and identifies the contour of a contiguous region.

Image Processing IP – Tensorflow Lite Inference

TF Lite Inference IP is a generic IP to implement Neural Network models on FPGA. In this case the TF Lite inference IP has been implemented for OCR and is available as an Overlay for OCR implementations.

Image Processing IP – Object Detection

Object Detection overlay is a readymade pipeline that combines multiple other IPs to compare features like shape, color and size of objects with a set of features that has been defined as a template.