I use mainly IV for viewing, and in batch mode, for renaming the pictures with meaningful names.
Often I would like to reduce the size, but keep the high quality definition of the main subject, while decreasing the quality of surrounding, especially if it has no artistic value or is not scenic.
I have read in Wikipedia how a picture is encoded and understood the few steps:
- Color space transformation,
- Down-sampling
- Block splitting
- Discrete cosine transform
- Quantization
- Entropy coding (compress)
The principle to avoid further loss of quality is to keep the data of each block after "Quantization" unchanged.
The steps could be:
- Load file
- Uncompress
- Restore data of each block (B matrix in Wikipedia) and keep it
- Continue process of recreating image
- Display image
- Change quality of some areas (this action can be repeated)
-- The user selects (polygon) an area
-- The software automatically selects full blocks corresponding to the one stored in the file.
-- The user defines a down-sampling for this area (can be horizontal, vertical or both directions)
-- The modified blocks are marked
- When saving is requested do for all marked modified blocks
-- Color space transformation
-- (No need for further down-sampling)
-- Discrete cosine transform
-- Quantization
After this operation, blocks have been modified, but blocks with no changes to the image will not be changed, thus leading to no further loss of quality to those ones.
- Compress
- Save file
The explanation may look long as I want to make sure that the request is well understood.
This function has following advantages:
- keep untouched the quality of picture of the main subject of the photo as delivered by the camera,
- reduce the quality (or sharpness) of secondary zone of the photo
- reduce the size of the file, without loss of viewer's satisfaction.
Thank you.
Often I would like to reduce the size, but keep the high quality definition of the main subject, while decreasing the quality of surrounding, especially if it has no artistic value or is not scenic.
I have read in Wikipedia how a picture is encoded and understood the few steps:
- Color space transformation,
- Down-sampling
- Block splitting
- Discrete cosine transform
- Quantization
- Entropy coding (compress)
The principle to avoid further loss of quality is to keep the data of each block after "Quantization" unchanged.
The steps could be:
- Load file
- Uncompress
- Restore data of each block (B matrix in Wikipedia) and keep it
- Continue process of recreating image
- Display image
- Change quality of some areas (this action can be repeated)
-- The user selects (polygon) an area
-- The software automatically selects full blocks corresponding to the one stored in the file.
-- The user defines a down-sampling for this area (can be horizontal, vertical or both directions)
-- The modified blocks are marked
- When saving is requested do for all marked modified blocks
-- Color space transformation
-- (No need for further down-sampling)
-- Discrete cosine transform
-- Quantization
After this operation, blocks have been modified, but blocks with no changes to the image will not be changed, thus leading to no further loss of quality to those ones.
- Compress
- Save file
The explanation may look long as I want to make sure that the request is well understood.
This function has following advantages:
- keep untouched the quality of picture of the main subject of the photo as delivered by the camera,
- reduce the quality (or sharpness) of secondary zone of the photo
- reduce the size of the file, without loss of viewer's satisfaction.
Thank you.
Comment