Image processing with minimization of ringing artifacts and noise
AssigneeTexas Instruments Incorporated
InventorsJeffrey Kempf, Arnold P. Skoog, Roger M. Ikeda
Life of this patent
Estimating the term…
Abstract
A method of reducing ringing artifacts in image data that has been filtered with a high frequency emphasis filter. For each filtered data value, a local variance is calculated from data values at neighboring filter taps. This variance is compared to a threshold, and if the threshold is exceeded, the filtered data value is limited between local minimum and maximum values. A method of reducing noise, also using the local variance, is also described.
Claims8
A circuit, comprising:
- a multi-tap filter for filtering a sequence of image data values, to produce a sequence of filtered data values;
- a maximum/minimum finder circuit for finding a pair of maximum/minimum values over a set of neighboring image data values for each filtered data value;
- a subtractor for calculating a local variance for each filtered data value, the local variance corresponding to the difference between the pair of maximum/minimum values,
- a comparator for comparing the local variance to a variance threshold;
- and
- a limiter, receiving the filtered data values, for limiting each filtered data value between maximum and minimum values for that data value responsive to the local variance for that filtered data value exceeding the variance threshold.
Circuitry for reducing noise in image data that is to be filtered by a filter having a number of filter taps and a number of filter coefficients, comprising:
- circuitry for determining a local variance corresponding to the difference between the unfiltered maximum and minimum values over a set of neighboring image data values;
- a first comparator for comparing the local variance to a first variance threshold;
- a second comparator for comparing the local variance to a second variance threshold;
- a multiplexer for selecting the filter coefficients as follows:
- responsive to the local variance is less than the first threshold, selecting filter coefficients to configure the filter as a linear filter;
- responsive to the local variance is between the first and the second threshold, selecting filter coefficients to configure the filter as a semi-linear filter;
- and
- responsive to the local variance is greater than both thresholds, selecting filter coefficients to configure the filter as a normal filter;
- and
- a multi-tap filter for filtering image data based on the selected filter coefficients.
Dependencies are inferred from the claim text.