Image processing operations, such as image convolution and mathematical morphology operations, are essential for filtering, enhancing, or detecting image features of interest. Performing the image processing operations is a computationally intensive task, especially in applications that require operations with large kernels and or many image processing steps . Hardware acceleration has been used by many machine vision systems to achieve the image processing throughput requirement.

The continued rapid advancement of software, computing, storage, database, communication, memory bandwidth, digital image acquisition, and streaming media technologies has created a paradigm shift in the machine vision approach. Specialized machine vision hardware accelerator and associated software technologies are becoming obsolete and general-purpose computing technologies are being widely adapted for machine vision applications. The drivers are low cost and flexible application software. Previously, low cost software solutions were restricted to simple or non-time critical machine vision applications.

DRVision Technologies LLC high-speed image processing technology converts a general purpose PC into a high-speed image processing computer. It does this without adding hardware accelerators because it had developed innovative software solutions to many sophisticated machine vision applications. DRVision Technologies LLC accomplishes high-speed image processing for filtering, enhancement, and detection through a sequenced cascade of elongated filter operations that produce the same or equivalent result as a large kernel operation but use very limited computing resources. The sum of the time required for the sequence of cascaded elongated filter operations is substantially less than what is required when a large kernel is employed to accomplish the same end result directly. Each elongated filter operation involves only a small one-dimensional kernel. The elongated filter kernels are simple and predictable. The simplicity and predictability improves the effectiveness of memory caching and operation predictions that are built into many CPUs. The speed of elongated filter processing is kernel size insensitive. Therefore, large kernel operations (programmed as a cascade of elongated filters) can be freely applied without significant impact to processing throughput.

Because of the fundamental algorithmic architectural innovation, DRVision Technologies LLC high-speed image processing technology enables real-time image processing programmed in software on general purpose CPUs to accomplish sophisticated machine vision applications.