Technologies
Precoad Video Noise Reduction (VNR)
Introductory Backgrounder
This introductory backgrounder provides guidance for greater understanding and
thereby getting the most out of the Precoad VNR “programmable” software
Evaluation Tool (ET).
The VNR Model
The Precoad VNR solution is an anisotropic-diffusion-based spatial-temporal VNR
solution. The anisotropic-diffusion-based model has been highly regarded for
over 2 decades as the best model in the world by universities, academic
institutions, governments, military institutions, professional societies, journals, and
forums. The special feature of anisotropic diffusion is its imagery feature
preservation, which overcomes the blurring and feature loss problems of the
isotropic-diffusion-based models that have been available for 3 decades.
The Solutions Attempted
The anisotropic-diffusion model is characterized by the anisotropic-diffusion partial
differential equation (PDE). This is a highly non-linear PDE that is extremely
difficult to solve numerically. The multiple-pass iterative (approximate) numerical
approaches that have been attempting to solve the PDE have proved prohibitively
complex, iteratively destructive in error propagation, heavy computationally and
resource intensive, non-real-time, off-line, slow and high-cost, basically non-
practical and non-implementable.
The Precoad Solution – the Innovation
Instead of solving the PDE numerically, the patented Precoad approach
introduces “pixel-dependent scales” which uniquely provide the Precoad VNR
solution as a single-pass filter. Technically speaking, this elegant mathematical
solution basically topologically “maps” to the “clean” target video in one-pass. As
a result, this thereby uniquely provides the lowest implementation cost, complexity,
processing, and cache memory, with enhanced video imagery quality. This is the
basis for the Precoad “heaviest-weight” VNR performance, in a “lightest-
weight” VNR engine.
This unique single-pass VNR solution is furthermore implemented in a low-memory
usage simple form which further provides for the live, real-time, speed execution of
the Precoad VNR solution using minimal resources.
The Precoad VNR “Programmable” Software Evaluation Tool (ET)
These simple forms are the basis of the various VNR strength levels and
parameter settings in the Precoad VNR Software ET, especially in the “Advanced”
tab of the ET. Each VNR strength level (corresponding to the parameter settings
in the tabs of the ET) corresponds to a simple form. These are furthermore
“programmable”, as these parameters and settings are user-programmable, and
the user parameter settings files can be simply saved and loaded into the ET.
The amazing demonstration of the “lightest-weight” benefit of the Precoad VNR
engine is that it is implemented in software ET – for live, real-time “heaviest-
weight” VNR performance, for resolutions up to at least full HD 1920x1080p, 30
frames per second.
“Behind the (ET) GUI”
User tuning, optimization, and customization is provided with the user-friendly ET.
ET provides a typical standard type video engineering design and development
GUI, particularly in the “Advanced” tab of the ET. This segments temporal and
spatial VNR, with cross-terms between temporal and spatial VNR. This also
segments across the Y, U, and V channels.
The heavy mathematical macro-level adjustment and tuning is done “behind the
GUI”. This entails the mathematical adjustment and adaptation of the “transition
parameter”, p0, of the anisotropic-diffusion PDE, for suitable balance between
feature preservation/imagery sharpening and video imagery smoothing. The ET
provides these in the form of 2 baseline adaptations -- one adaptation for
“Security camera/Video camcorder” type applications, and the other adaptation
for LCD/PDP HDTV type applications.
FPGA Demo Implementation of VNR Core Engine
The Precoad VNR Core engine has been implemented in an FPGA demo, in
matching Precoad VNR Core FPGA Synthesizable RTL Verilog-HDL Code.
Precoad Video Noise Reduction (VNR)
Introductory Backgrounder
This introductory backgrounder provides guidance for greater understanding and
thereby getting the most out of the Precoad VNR “programmable” software
Evaluation Tool (ET).
The VNR Model
The Precoad VNR solution is an anisotropic-diffusion-based spatial-temporal VNR
solution. The anisotropic-diffusion-based model has been highly regarded for
over 2 decades as the best model in the world by universities, academic
institutions, governments, military institutions, professional societies, journals, and
forums. The special feature of anisotropic diffusion is its imagery feature
preservation, which overcomes the blurring and feature loss problems of the
isotropic-diffusion-based models that have been available for 3 decades.
The Solutions Attempted
The anisotropic-diffusion model is characterized by the anisotropic-diffusion partial
differential equation (PDE). This is a highly non-linear PDE that is extremely
difficult to solve numerically. The multiple-pass iterative (approximate) numerical
approaches that have been attempting to solve the PDE have proved prohibitively
complex, iteratively destructive in error propagation, heavy computationally and
resource intensive, non-real-time, off-line, slow and high-cost, basically non-
practical and non-implementable.
The Precoad Solution – the Innovation
Instead of solving the PDE numerically, the patented Precoad approach
introduces “pixel-dependent scales” which uniquely provide the Precoad VNR
solution as a single-pass filter. Technically speaking, this elegant mathematical
solution basically topologically “maps” to the “clean” target video in one-pass. As
a result, this thereby uniquely provides the lowest implementation cost, complexity,
processing, and cache memory, with enhanced video imagery quality. This is the
basis for the Precoad “heaviest-weight” VNR performance, in a “lightest-
weight” VNR engine.
This unique single-pass VNR solution is furthermore implemented in a low-memory
usage simple form which further provides for the live, real-time, speed execution of
the Precoad VNR solution using minimal resources.
The Precoad VNR “Programmable” Software Evaluation Tool (ET)
These simple forms are the basis of the various VNR strength levels and
parameter settings in the Precoad VNR Software ET, especially in the “Advanced”
tab of the ET. Each VNR strength level (corresponding to the parameter settings
in the tabs of the ET) corresponds to a simple form. These are furthermore
“programmable”, as these parameters and settings are user-programmable, and
the user parameter settings files can be simply saved and loaded into the ET.
The amazing demonstration of the “lightest-weight” benefit of the Precoad VNR
engine is that it is implemented in software ET – for live, real-time “heaviest-
weight” VNR performance, for resolutions up to at least full HD 1920x1080p, 30
frames per second.
“Behind the (ET) GUI”
User tuning, optimization, and customization is provided with the user-friendly ET.
ET provides a typical standard type video engineering design and development
GUI, particularly in the “Advanced” tab of the ET. This segments temporal and
spatial VNR, with cross-terms between temporal and spatial VNR. This also
segments across the Y, U, and V channels.
The heavy mathematical macro-level adjustment and tuning is done “behind the
GUI”. This entails the mathematical adjustment and adaptation of the “transition
parameter”, p0, of the anisotropic-diffusion PDE, for suitable balance between
feature preservation/imagery sharpening and video imagery smoothing. The ET
provides these in the form of 2 baseline adaptations -- one adaptation for
“Security camera/Video camcorder” type applications, and the other adaptation
for LCD/PDP HDTV type applications.
FPGA Demo Implementation of VNR Core Engine
The Precoad VNR Core engine has been implemented in an FPGA demo, in
matching Precoad VNR Core FPGA Synthesizable RTL Verilog-HDL Code.
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
