top of page

Lemur Imaging's LMR can be used for both Visual and Computer Vision applications 

Camera Pipeline 

LMR IP can be instantiated anywhere where line memory is used.  
There are no cumulative encoding/decoding artefacts and the cores can safely be instantiated multiple times across different convolutions.  All data formats and precisions are supported including planar, semi planar and all HDR formats. 

Display Pipeline

LMR IP can be used to reduce the size of delay lines associated with scalers, FP16 data format and higher screen resolutions.  

H.264/HEVC Encode

LMR IP can be used to reduce the size of delay lines associated with H.264/HEVC video encoders with no impact to final quality of encoded frames. All encoded frames continue to adhere to H.264/HEVC encoding standards and can be decoded with no issue by all H.264/HEVC decoders. 

Typical Convolution filter 

Convolution filiter - no LMR_edited.png

Convolution filter with LMR reducing line memory 

Convolution filter with LMR_edited.jpg

CV Engine

LMR IP can reduce the delay lines associated with feature compression, with no impact on the quality of the end detections. It can be used with 8bit data and FP16 data.  
LMR+FP16 data implementation can be instantiated in a smaller silicon area than FP8 - whilst delivering more accurate detections. 

Screenshot 2022-08-31 at 16.45.34.png
bottom of page