Adding comments at each method, referencing the concerning articles.
--- a/src/jpeg-6b/jdrfrom.c Mon Sep 18 01:29:41 2006 +0200
+++ b/src/jpeg-6b/jdrfrom.c Wed Oct 04 20:46:11 2006 +0200
@@ -6,6 +6,16 @@
#include "jdrest.h"
+/* Implementation of the method developed by: François Alter, Sylvain Durand and
+ Jacques Froment:
+ Adapted Total Variation for Artifact Free Decompression of JPEG Images
+ from Journal of Mathematical Imaging and Vision 23: 199-211, 2005
+
+ Notes:
+ - Big part of the implementation has been freely given by Jacques Froment.
+ - This method (or implementation) noticeable decreases the image contrast.
+*/
+
#define MAXSAMP 255.
#define MINSAMP 0.
--- a/src/jpeg-6b/jdrhp.c Mon Sep 18 01:29:41 2006 +0200
+++ b/src/jpeg-6b/jdrhp.c Wed Oct 04 20:46:11 2006 +0200
@@ -5,6 +5,19 @@
#include "jpeglib.h"
#include "jdrest.h"
+/* Implementation of the method developed by Ramin Samadani, Arvind
+ * Sundararajan and Amir Said:
+ * Deringing and deblocking DCT compression artifacts with efficient
+ * shifted transforms.
+ * from the online version published at HP's site.
+ *
+ * Notes:
+ * - We implement the version which uses DCTs/iDCTs. Basically a variance
+ * over Nosratinia's method, locally choosing the shifts to average for
+ * obtaining the final point.
+ * - This methods takes _a lot of memory_, around 65 times each image pixel,
+ * with a float for every pixel.
+ */
#define MAXSAMP 255.
#define MINSAMP 0.
--- a/src/jpeg-6b/jdrnosr.c Mon Sep 18 01:29:41 2006 +0200
+++ b/src/jpeg-6b/jdrnosr.c Wed Oct 04 20:46:11 2006 +0200
@@ -4,6 +4,14 @@
#include "jpeglib.h"
#include "jdrest.h"
+/* Implementation of the method developed by Aria Nosratinia:
+ Enhancement of JPEG-Compressed Images by Re-application of JPEG
+ (c) 2002 Kluwer Academic Publishers.
+
+ Notes:
+ - The implementation from Nikhil Hegde has been a reference for writting
+ this one.
+*/
#define MAXSAMP 255.
#define MINSAMP 0.
--- a/src/jpeg-6b/jdrrober.c Mon Sep 18 01:29:41 2006 +0200
+++ b/src/jpeg-6b/jdrrober.c Wed Oct 04 20:46:11 2006 +0200
@@ -3,20 +3,17 @@
#include "jpeglib.h"
#include "jdrest.h"
-/* Implementation of the method developed by Thomas P. O'Rourke and
- * Robert L. Stevenson in 1995 (IEEE !!!!??!!)
+/* Implementation of the method developed by Mark A. Robertson and
+ * Robert L. Stevenson:
+ * DCT Quantization Noise in Compressed Images
+ * draft published online, dated of February 23rd, 2004.
*
- * There are a few changes over the original:
- * - The gradient is calculated as:
- * g^(k) = - sum(c in C, rho_T' (d_c^t * z^(k)) * d_c)
- * instead of
- * g^(k) = sum(c in C, rho_T' (d_c^t * z^(k)) * d_c^t)
- * - The step size isn't calculated. It's tried from 1, and halving
- * as much as needed for minimizing the objective function
- * sum(c in C, rho_T(z_{m,n} - z_{k,l}))
- * - The step size tried has a minimum on halving (MINIMUM_STEP_SIZE). Having
- * reached the minimum, no more steps are done, as they wouldn't improve
- * the bitmap according to the method constraints.
+ * Notes:
+ * - The article doesn't mention how to find a good \alpha^{(w)}.
+ * We start from an alpha equal to the proposed step size, and then
+ * halving until a good candidate is found (decreasing the term that we have
+ * to minimize). At MINIMUM_STEP_SIZE, we stop trying to minimize, and exit
+ * the minimizing loop.
*/
#define MAXSAMP 255.
--- a/src/jpeg-6b/jdrrour.c Mon Sep 18 01:29:41 2006 +0200
+++ b/src/jpeg-6b/jdrrour.c Wed Oct 04 20:46:11 2006 +0200
@@ -4,7 +4,10 @@
#include "jdrest.h"
/* Implementation of the method developed by Thomas P. O'Rourke and
- * Robert L. Stevenson in 1995 (IEEE !!!!??!!)
+ * Robert L. Stevenson:
+ * Improved Image Decompression for Reduced Transform Coding Artifacts
+ * from IEEE Transactions on Circuits and Systems for Video Technology,
+ * vol. 5, No. 6, December 1995.
*
* There are a few changes over the original:
* - The gradient is calculated as: