add a -partition_limit option to limit the number of bits used by intra4x4

Although it degrades quality, this option is useful to avoid the 512k
limit for partition #0.
If not enough to reach the lower bound of 4bits per macroblock header,
one should also limit the number of segments used (down to -segments 1)

See the man file for extra details.

Change-Id: Ia59ffac13176c85b809ddd6340d37b54ee9487ea

src/enc/config.c

@@ -42,6 +42,7 @@ int WebPConfigInitInternal(WebPConfig* const config,
   config->preprocessing = 0;
   config->autofilter = 0;
   config->alpha_compression = 0;
+  config->partition_limit = 0;
 
   // TODO(skal): tune.
   switch (preset) {
@@ -106,6 +107,8 @@ int WebPValidateConfig(const WebPConfig* const config) {
     return 0;
   if (config->partitions < 0 || config->partitions > 3)
     return 0;
+  if (config->partition_limit < 0 || config->partition_limit > 100)
+    return 0;
   if (config->alpha_compression < 0)
     return 0;
   return 1;

src/enc/quant.c

@@ -757,8 +757,13 @@ static int PickBestIntra4(VP8EncIterator* const it, VP8ModeScore* const rd) {
   const int tlambda = dqm->tlambda_;
   const uint8_t* const src0 = it->yuv_in_ + Y_OFF;
   uint8_t* const best_blocks = it->yuv_out2_ + Y_OFF;
+  int total_header_bits = 0;
   VP8ModeScore rd_best;
 
+  if (enc->max_i4_header_bits_ == 0) {
+    return 0;
+  }
+
   InitScore(&rd_best);
   rd_best.score = 211;  // '211' is the value of VP8BitCost(0, 145)
   VP8IteratorStartI4(it);
@@ -799,7 +804,9 @@ static int PickBestIntra4(VP8EncIterator* const it, VP8ModeScore* const rd) {
     }
     SetRDScore(dqm->lambda_mode_, &rd_i4);
     AddScore(&rd_best, &rd_i4);
-    if (rd_best.score >= rd->score) {
+    total_header_bits += mode_costs[best_mode];
+    if (rd_best.score >= rd->score ||
+        total_header_bits > enc->max_i4_header_bits_) {
       return 0;
     }
     // Copy selected samples if not in the right place already.

src/enc/vp8enci.h

@@ -359,8 +359,9 @@ struct VP8Encoder {
   int      block_count_[3];
 
   // quality/speed settings
-  int method_;             // 0=fastest, 6=best/slowest.
-  int rd_opt_level_;       // Deduced from method_.
+  int method_;              // 0=fastest, 6=best/slowest.
+  int rd_opt_level_;        // Deduced from method_.
+  int max_i4_header_bits_;  // partition #0 safeness factor
 
   // Memory
   VP8MBInfo* mb_info_;   // contextual macroblock infos (mb_w_ + 1)

src/enc/webpenc.c

@@ -112,11 +112,15 @@ static void ResetBoundaryPredictions(VP8Encoder* const enc) {
 
 static void MapConfigToTools(VP8Encoder* const enc) {
   const int method = enc->config_->method;
+  const int limit = 100 - enc->config_->partition_limit;
   enc->method_ = method;
   enc->rd_opt_level_ = (method >= 6) ? 3
                      : (method >= 5) ? 2
                      : (method >= 3) ? 1
                      : 0;
+  enc->max_i4_header_bits_ =
+      256 * 16 * 16 *                 // upper bound: up to 16bit per 4x4 block
+      (limit * limit) / (100 * 100);  // ... modulated with a quadratic curve.
 }
 
 // Memory scaling with dimensions:

src/webp/encode.h

@@ -69,6 +69,8 @@ typedef struct {
   int preprocessing;     // preprocessing filter (0=none, 1=segment-smooth)
   int partitions;        // log2(number of token partitions) in [0..3]
                          // Default is set to 0 for easier progressive decoding.
+  int partition_limit;   // quality degradation allowed to fit the 512k limit on
+                         // prediction modes coding (0=no degradation, 100=full)
   int alpha_compression;  // Algorithm for optimizing the alpha plane (0 = none)
 } WebPConfig;