An FPGA Implementation for Texture Analysis Considering the Real-Time Requirements of Vision-Based Systems

Analysis of the efficiency of the census transform algorithm implemented on FPGA

“…Census Transform-based implementations have been well adapted particularly for obstacle detection and mobile robotics applications as shown in [15]. In the cited work an implementation of an optimized architecture based on Census Transform is presented, as well as a performance comparison between three different devices: FPGA, DSP and PC for three different Census window sizes.…”

“…In [16] a similar architecture is designed using high-level synthesis software (GAUT), and a comparison among other works in terms of the resources needed for implementation is described. The architectures presented in [15,16] are then applied to an obstacle detection application on a mobile robot [1].…”

“…This requires that all the pixels that will be used in a clock cycle be stored in registers and the remaining streamed (but unused) pixels be stored in temporary memory buffers. For the hardware implementation we use the architecture presented in [15] as basis for our design of the parameter scalable architecture. The image input is considered of W  H (Width  Height) pixels and the pixels are of nb bits width.…”

Fast texture evaluation of textiles using the GLBP technique in GPU architecture

“…15 Extraction of texture features for real-time applications can be accelerated by implementing algorithms, based on exploring local texture using repetitive functions i.e co-occurrence matrix and Local Binary Pattern (LBP) techniques, in parallel programming environments. [16][17][18][19] Particularly, LBP techniques have proven to be efficient for texture recognition and classification tasks while demand low computational cost. 20,21 Additionally, this type of techniques have also the advantage over other techniques in terms of tolerance against illumination variations.…”