diff --git a/ASTCEncoder/src/Decompressor.cpp b/ASTCEncoder/src/Decompressor.cpp
index afcb0fb..5b2f1cf 100644
--- a/ASTCEncoder/src/Decompressor.cpp
+++ b/ASTCEncoder/src/Decompressor.cpp
@@ -605,7 +605,7 @@ namespace ASTCC {
     uint32 Ds = (1024 + (blockWidth/2)) / (blockWidth - 1);
     uint32 Dt = (1024 + (blockHeight/2)) / (blockHeight - 1);
 
-    for(uint32 plane = 0; plane < (params.m_bDualPlane? 2 : 1); plane++)
+    for(uint32 plane = 0; plane < (params.m_bDualPlane? 2U : 1U); plane++)
     for(uint32 t = 0; t < blockHeight; t++)
     for(uint32 s = 0; s < blockWidth; s++) {
       uint32 cs = Ds * s;
diff --git a/BPTCEncoder/src/CompressNVTT.cpp b/BPTCEncoder/src/CompressNVTT.cpp
index e07bda1..d7737e9 100644
--- a/BPTCEncoder/src/CompressNVTT.cpp
+++ b/BPTCEncoder/src/CompressNVTT.cpp
@@ -112,17 +112,16 @@ namespace BPTCC {
     uint8 *outBuf = cj.OutBuf() + cj.CoordsToBlockIdx(cj.XStart(), cj.YStart()) * kBlockSz;
 
     uint32 startX = cj.XStart();
-    bool done = false;
-
-    for(uint32 j = cj.YStart(); !done; j += 4) {
-      for(uint32 i = startX; !done && i < cj.Width(); i += 4) {
+    const uint32 endY = std::min(cj.YEnd(), cj.Height() - 4);
+    for(uint32 j = cj.YStart(); j <= endY; j += 4) {
+      const uint32 endX = j == cj.YEnd()? cj.XEnd() : cj.Width();
+      for(uint32 i = startX; i < endX; i += 4) {
 
         Tile block(4, 4);
         GetBlock(i, j, cj.Width(), inPixels, block);
         AVPCL::compress(block, reinterpret_cast<char *>(outBuf), NULL);
 
         outBuf += kBlockSz;
-        done = i+4 >= cj.XEnd() && j+(i+4 == cj.Width()? 4 : 0) >= cj.YEnd();
       }
       startX = 0;
     }
@@ -173,9 +172,10 @@ namespace BPTCC {
     uint8 *outBuf = cj.OutBuf() + cj.CoordsToBlockIdx(cj.XStart(), cj.YStart()) * kBlockSz;
 
     uint32 startX = cj.XStart();
-    bool done = false;
-    for(uint32 j = cj.YStart(); !done; j += 4) {
-      for(uint32 i = startX; !done && i < cj.Width(); i += 4) {
+    const uint32 endY = std::min(cj.YEnd(), cj.Height() - 4);
+    for(uint32 j = cj.YStart(); j <= endY; j += 4) {
+      const uint32 endX = j == cj.YEnd()? cj.XEnd() : cj.Width();
+      for(uint32 i = startX; i < endX; i += 4) {
 
         Tile block(4, 4);
         GetBlock(i, j, cj.Width(), inPixels, block);
@@ -197,7 +197,6 @@ namespace BPTCC {
         }
 
         outBuf += 16;
-        done = i+4 >= cj.XEnd() && j+(i+4 == cj.Width()? 4 : 0) >= cj.YEnd();
       }
 
       startX = 0;
diff --git a/BPTCEncoder/src/Compressor.cpp b/BPTCEncoder/src/Compressor.cpp
index 0d2ff6b..2c8d928 100755
--- a/BPTCEncoder/src/Compressor.cpp
+++ b/BPTCEncoder/src/Compressor.cpp
@@ -1572,8 +1572,9 @@ void Compress(const FasTC::CompressionJob &cj, CompressionSettings settings) {
   const uint32 kBlockSz = GetBlockSize(FasTC::eCompressionFormat_BPTC);
   uint8 *outBuf = cj.OutBuf() + cj.CoordsToBlockIdx(cj.XStart(), cj.YStart()) * kBlockSz;
 
+  const uint32 endY = std::min(cj.YEnd(), cj.Height() - 4);
   uint32 startX = cj.XStart();
-  for(uint32 j = cj.YStart(); j <= cj.YEnd(); j += 4) {
+  for(uint32 j = cj.YStart(); j <= endY; j += 4) {
     const uint32 endX = j == cj.YEnd()? cj.XEnd() : cj.Width();
     for(uint32 i = startX; i < endX; i += 4) {
 
diff --git a/BPTCEncoder/src/RGBAEndpoints.cpp b/BPTCEncoder/src/RGBAEndpoints.cpp
index f5aa349..18041da 100755
--- a/BPTCEncoder/src/RGBAEndpoints.cpp
+++ b/BPTCEncoder/src/RGBAEndpoints.cpp
@@ -79,9 +79,10 @@
 #include "CompressionMode.h"
 
 #include <cassert>
+#include <cfloat>
+#include <cmath>
 #include <cstdlib>
 #include <cstdio>
-#include <cfloat>
 
 #ifndef min
 template <typename T>
@@ -228,7 +229,7 @@ double RGBACluster::QuantizedError(
 ) const {
 
   // nBuckets should be a power of two.
-  const uint8 indexPrec = log2(nBuckets);
+  const uint8 indexPrec = static_cast<uint8>(log(static_cast<float>(nBuckets))/log(2.0f));
   assert(!(nBuckets & (nBuckets - 1)));
   assert(indexPrec >= 2 && indexPrec <= 4);
   
@@ -293,8 +294,8 @@ double RGBACluster::QuantizedError(
     const int32 j2 = static_cast<int32>(pct * static_cast<float>(nBuckets-1) + 0.7);
 #else
     const float pct = ((pt - uqp1) * uqpdir) / uqplsq;
-    int32 j1 = floor(pct * static_cast<float>(nBuckets-1));
-    int32 j2 = ceil(pct * static_cast<float>(nBuckets-1));
+    int32 j1 = static_cast<int32>(floor(pct * static_cast<float>(nBuckets-1)));
+    int32 j2 = static_cast<int32>(ceil(pct * static_cast<float>(nBuckets-1)));
     j1 = std::min(std::max(0, j1), nBuckets - 1);
     j2 = std::min(j2, nBuckets - 1);
 #endif
diff --git a/Base/include/Bits.h b/Base/include/Bits.h
index e593b8e..58138d9 100644
--- a/Base/include/Bits.h
+++ b/Base/include/Bits.h
@@ -70,8 +70,8 @@ namespace FasTC {
    public:
     explicit Bits(IntType &v) : m_Bits(v) { }
 
-    bool operator[](uint32 bitPos) {
-      return (m_Bits >> bitPos) & 1;
+    uint8 operator[](uint32 bitPos) {
+      return static_cast<uint8>((m_Bits >> bitPos) & 1);
     }
 
     IntType operator()(uint32 start, uint32 end) {
diff --git a/Base/include/MatrixSquare.h b/Base/include/MatrixSquare.h
index 3270981..c11fb33 100644
--- a/Base/include/MatrixSquare.h
+++ b/Base/include/MatrixSquare.h
@@ -55,7 +55,7 @@ namespace FasTC {
       int numIterations = 0;
 
       VectorBase<T, N> b;
-      T norm = 1.0/sqrt(static_cast<T>(N));
+      T norm = static_cast<T>(1.0)/sqrt(static_cast<T>(N));
       for(int i = 0; i < N; i++)
         b[i] = norm;
 
diff --git a/Base/src/Image.cpp b/Base/src/Image.cpp
index ea18b36..7240e7a 100644
--- a/Base/src/Image.cpp
+++ b/Base/src/Image.cpp
@@ -316,6 +316,11 @@ double Image<PixelType>::ComputeSSIM(Image<PixelType> *other) {
   const uint32 filterSz = 11;
   const double filterSigma = 1.5;
 
+  if(img1.GetWidth() < filterSz || img1.GetHeight() < filterSz ||
+     img2.GetWidth() < filterSz || img2.GetHeight() < filterSz) {
+    return -1.0;
+  }
+
   Image<IPixel> mu1 = FilterValid(img1, filterSz, filterSigma);
   Image<IPixel> mu2 = FilterValid(img2, filterSz, filterSigma);
 
diff --git a/Base/test/TestPixel.cpp b/Base/test/TestPixel.cpp
index 59c50da..71a3be3 100644
--- a/Base/test/TestPixel.cpp
+++ b/Base/test/TestPixel.cpp
@@ -82,7 +82,7 @@ TEST(Pixel, FourWideConstructor) {
     EXPECT_EQ(depth[i], 8);
   }
 
-  FasTC::Pixel q(static_cast<uint16>(1 << 16), 6, -2, 5, 4);
+  FasTC::Pixel q(0, 6, -2, 5, 4);
   EXPECT_EQ(q.R(), 6);
   EXPECT_EQ(q.G(), -2);
   EXPECT_EQ(q.B(), 5);
diff --git a/DXTEncoder/src/DXTCompressor.cpp b/DXTEncoder/src/DXTCompressor.cpp
index c98bc80..2f1ccd1 100755
--- a/DXTEncoder/src/DXTCompressor.cpp
+++ b/DXTEncoder/src/DXTCompressor.cpp
@@ -14,6 +14,7 @@
 // algorithms used in this code.
 
 #include "DXTCompressor.h"
+#include <algorithm>
 #include <cassert>
 #include <cstdlib>
 #include <cstring>
@@ -45,8 +46,9 @@ namespace DXTC
     uint8 *outBuf = cj.OutBuf() + startBlock * kBlockSz;
 
     const uint32 *inPixels = reinterpret_cast<const uint32 *>(cj.InBuf());
+    uint32 endY = std::min(cj.YEnd(), cj.Height() - 4);
     uint32 startX = cj.XStart();
-    for(uint32 j = cj.YStart(); j <= cj.YEnd(); j += 4) {
+    for(uint32 j = cj.YStart(); j <= endY; j += 4) {
       const uint32 endX = j == cj.YEnd()? cj.XEnd() : cj.Width();
       for(uint32 i = startX; i < endX; i += 4) {
 
@@ -75,10 +77,11 @@ namespace DXTC
     uint8 *outBuf = cj.OutBuf() + startBlock * kBlockSz;
     
     const uint32 *inPixels = reinterpret_cast<const uint32 *>(cj.InBuf());
+    uint32 endY = std::min(cj.YEnd(), cj.Height() - 4);
     uint32 startX = cj.XStart();
-    bool done = false;
-    for(uint32 j = cj.YStart(); !done; j += 4) {
-      for(uint32 i = startX; !done && i < cj.Width(); i += 4) {
+    for(uint32 j = cj.YStart(); j <= endY; j += 4) {
+      const uint32 endX = j == cj.YEnd()? cj.XEnd() : cj.Width();
+      for(uint32 i = startX; i < endX; i += 4) {
 
         const uint32 kOffset = j*cj.Width() + i;
         ExtractBlock(inPixels + kOffset, cj.Width(), block);
@@ -89,8 +92,8 @@ namespace DXTC
         EmitWord(outBuf, ColorTo565(maxColor));
         EmitWord(outBuf, ColorTo565(minColor));
         EmitColorIndices(block, outBuf, minColor, maxColor);
-        done = i+4 >= cj.XEnd() && j+(i+4 == cj.Width()? 4 : 0) >= cj.YEnd();
       }
+      startX = 0;
     }
   }
 
diff --git a/ETCEncoder/src/Compressor.cpp b/ETCEncoder/src/Compressor.cpp
index 90b4596..5afbb14 100644
--- a/ETCEncoder/src/Compressor.cpp
+++ b/ETCEncoder/src/Compressor.cpp
@@ -52,6 +52,7 @@
 
 #include "rg_etc1.h"
 #include "ETCCompressor.h"
+#include <algorithm>
 #include <cstring>
 
 namespace ETCC {
@@ -62,11 +63,13 @@ namespace ETCC {
     params.m_quality = rg_etc1::cLowQuality;
     rg_etc1::pack_etc1_block_init();
 
-    const uint32 kBlockSz = GetBlockSize(FasTC::eCompressionFormat_ETC1);
+    uint32 kBlockSz = GetBlockSize(FasTC::eCompressionFormat_ETC1);
     const uint32 startBlock = cj.CoordsToBlockIdx(cj.XStart(), cj.YStart());
     uint8 *outBuf = cj.OutBuf() + startBlock * kBlockSz;
+
+    const uint32 endY = std::min(cj.YEnd(), cj.Height() - 4);
     uint32 startX = cj.XStart();
-    for(uint32 j = cj.YStart(); j <= cj.YEnd(); j += 4) {
+    for(uint32 j = cj.YStart(); j <= endY; j += 4) {
       const uint32 endX = j == cj.YEnd()? cj.XEnd() : cj.Width();
       for(uint32 i = startX; i < endX; i += 4) {
 
diff --git a/IO/src/ImageFile.cpp b/IO/src/ImageFile.cpp
index a8619fd..36298c1 100644
--- a/IO/src/ImageFile.cpp
+++ b/IO/src/ImageFile.cpp
@@ -131,7 +131,7 @@ ImageFile::~ImageFile() {
   }
 
   if(m_FileData) {
-    delete m_FileData;
+    delete [] m_FileData;
     m_FileData = NULL;
   }
 }
@@ -144,7 +144,7 @@ bool ImageFile::Load() {
   }
   
   if(m_FileData) {
-    delete m_FileData;
+    delete [] m_FileData;
     m_FileData = NULL;
   }
 
diff --git a/PVRTCEncoder/CMakeLists.txt b/PVRTCEncoder/CMakeLists.txt
index 555c1dd..026222b 100644
--- a/PVRTCEncoder/CMakeLists.txt
+++ b/PVRTCEncoder/CMakeLists.txt
@@ -68,6 +68,8 @@ SET( SOURCES
   src/PVRTCImage.cpp
 )
 
+OPTION(DEBUG_PVRTC_DECODER "Output intermediate images during PVRTC decoding." OFF)
+
 CONFIGURE_FILE(
   "config/PVRTCDefines.h.in"
   "include/PVRTCDefines.h"
@@ -87,7 +89,9 @@ ADD_LIBRARY( PVRTCEncoder
 )
 
 TARGET_LINK_LIBRARIES( PVRTCEncoder FasTCBase )
-TARGET_LINK_LIBRARIES( PVRTCEncoder FasTCIO )
+IF( DEBUG_PVRTC_DECODER )
+  TARGET_LINK_LIBRARIES( PVRTCEncoder FasTCIO )
+ENDIF( DEBUG_PVRTC_DECODER )
 
 IF( PVRTEXLIB_FOUND )
   TARGET_LINK_LIBRARIES( PVRTCEncoder ${PVRTEXLIB_LIBRARIES} )
diff --git a/PVRTCEncoder/config/PVRTCDefines.h.in b/PVRTCEncoder/config/PVRTCDefines.h.in
index 9aabc8d..d5aaa5f 100644
--- a/PVRTCEncoder/config/PVRTCDefines.h.in
+++ b/PVRTCEncoder/config/PVRTCDefines.h.in
@@ -51,3 +51,4 @@
  */
 
 #cmakedefine PVRTEXLIB_FOUND
+#cmakedefine DEBUG_PVRTC_DECODER
diff --git a/PVRTCEncoder/src/PVRTCImage.cpp b/PVRTCEncoder/src/PVRTCImage.cpp
index 2300176..a3c1f2d 100644
--- a/PVRTCEncoder/src/PVRTCImage.cpp
+++ b/PVRTCEncoder/src/PVRTCImage.cpp
@@ -66,7 +66,9 @@
 #include "Pixel.h"
 using FasTC::Pixel;
 
-#include "../../IO/include/ImageFile.h"
+#ifdef DEBUG_PVRTC_DECODER
+#  include "../../IO/include/ImageFile.h"
+#endif
 
 template <typename T>
 inline T Clamp(const T &v, const T &a, const T &b) {
@@ -540,6 +542,7 @@ uint32 Image::GetPixelIndex(int32 i, int32 j, EWrapMode wrapMode) const {
   return idx;
 }
 
+#ifdef DEBUG_PVRTC_DECODER
 void Image::DebugOutput(const char *filename) const {
   uint32 *outPixels = new uint32[GetWidth() * GetHeight()];
   const uint8 fullDepth[4] = { 8, 8, 8, 8 };
@@ -561,5 +564,8 @@ void Image::DebugOutput(const char *filename) const {
   ::ImageFile imgFile(debugFilename, eFileFormat_PNG, img);
   imgFile.Write();
 }
+#else
+void Image::DebugOutput(const char *filename) const { }
+#endif // DEBUG_PVRTC_DECODER
 
 }  // namespace PVRTCC
diff --git a/PVRTCEncoder/test/CMakeLists.txt b/PVRTCEncoder/test/CMakeLists.txt
index a399ee8..74190e0 100644
--- a/PVRTCEncoder/test/CMakeLists.txt
+++ b/PVRTCEncoder/test/CMakeLists.txt
@@ -56,6 +56,11 @@ INCLUDE_DIRECTORIES(${FasTC_SOURCE_DIR}/PVRTCEncoder/src)
 INCLUDE_DIRECTORIES(${FasTC_SOURCE_DIR}/Base/include )
 INCLUDE_DIRECTORIES(${FasTC_BINARY_DIR}/Base/include )
 
+IF( DEBUG_PVRTC_DECODER )
+  INCLUDE_DIRECTORIES(${FasTC_SOURCE_DIR}/Core/include )
+  INCLUDE_DIRECTORIES(${FasTC_SOURCE_DIR}/IO/include )
+ENDIF( DEBUG_PVRTC_DECODER )
+
 INCLUDE_DIRECTORIES(${FasTC_SOURCE_DIR}/GTest/include)
 
 SET(TESTS
diff --git a/PVRTCEncoder/test/DecompTestPVR.cpp b/PVRTCEncoder/test/DecompTestPVR.cpp
index af909d9..599ac0f 100644
--- a/PVRTCEncoder/test/DecompTestPVR.cpp
+++ b/PVRTCEncoder/test/DecompTestPVR.cpp
@@ -60,12 +60,10 @@
 
 #include "TestUtils.h"
 
-// #define OUTPUT_DEBUG_IMAGE
-
-#ifdef OUTPUT_DEBUG_IMAGE
-#  include "Core/include/Image.h"
-#  include "IO/include/ImageFile.h"
-#endif  // OUTPUT_DEBUG_IMAGE
+#ifdef DEBUG_PVRTC_DECODER
+#  include "Image.h"
+#  include "ImageFile.h"
+#endif  // DEBUG_PVRTC_DECODER
 
 class ImageTester {
  public:
@@ -88,7 +86,7 @@ class ImageTester {
 
     uint8 *outBuf = reinterpret_cast<uint8 *>(outPixels);
     FasTC::DecompressionJob dcj(fmt, data, outBuf, w, h);
-#ifdef OUTPUT_DEBUG_IMAGE
+#ifdef DEBUG_PVRTC_DECODER
     PVRTCC::Decompress(dcj, PVRTCC::eWrapMode_Wrap, true);
 #else
     PVRTCC::Decompress(dcj, PVRTCC::eWrapMode_Wrap);
@@ -106,10 +104,10 @@ class ImageTester {
       EXPECT_EQ(PixelPrinter(libPixels[i]), PixelPrinter(outPixels[i]));
     }
 
-#ifdef OUTPUT_DEBUG_IMAGE
+#ifdef DEBUG_PVRTC_DECODER
     char dbgfname[256];
     snprintf(dbgfname, sizeof(dbgfname), "Debug%s.png", filename);
-    ::ImageFile imgFile(dbgfname, eFileFormat_PNG, ::Image(w, h, outPixels));
+    ImageFile imgFile(dbgfname, eFileFormat_PNG, FasTC::Image<>(w, h, outPixels));
     imgFile.Write();
 #endif  // OUTPUT_DEBUG_IMAGE