diff --git a/ext/decoders/native/main.c b/ext/decoders/native/main.c
index 77f541f..90829d7 100644
--- a/ext/decoders/native/main.c
+++ b/ext/decoders/native/main.c
@@ -2,6 +2,15 @@
 #include <stdint.h>
 #include <ruby.h>
 #include "astc.h"
+#include "rgb.h"
+
+static VALUE rb_decode_rgb565(VALUE self, VALUE rb_data, VALUE size, VALUE big) {
+    uint8_t *image = (uint8_t*)malloc(FIX2LONG(size) * 3);
+    decode_rgb565((uint16_t*)RSTRING_PTR(rb_data), FIX2INT(size), RTEST(big), image);
+    VALUE ret = rb_str_new((char*)image, FIX2LONG(size) * 3);
+    free(image);
+    return ret;
+}
 
 static VALUE rb_decode_astc(VALUE self, VALUE rb_data, VALUE w, VALUE h, VALUE bw, VALUE bh) {
     const uint8_t *data = (uint8_t*)RSTRING_PTR(rb_data);
@@ -16,4 +25,5 @@ void Init_native() {
     VALUE mMikunyan = rb_define_module("Mikunyan");
     VALUE mDecodeHelper = rb_define_module_under(mMikunyan, "DecodeHelper");
 	rb_define_module_function(mDecodeHelper, "decode_astc", rb_decode_astc, 5);
+    rb_define_module_function(mDecodeHelper, "decode_rgb565", rb_decode_rgb565, 3);
 }
diff --git a/ext/decoders/native/rgb.c b/ext/decoders/native/rgb.c
new file mode 100644
index 0000000..6d4f7e6
--- /dev/null
+++ b/ext/decoders/native/rgb.c
@@ -0,0 +1,30 @@
+#include <stdint.h>
+
+static inline int is_system_little() {
+    int x = 1;
+    return *(char*)&x == 1;
+}
+
+void decode_rgb565(const uint16_t* data, const int size, const int is_big_endian, uint8_t* image) {
+    const uint16_t *d = data;
+    uint8_t *p = image;
+    if (is_big_endian == is_system_little()) {
+        for (int i = 0; i < size; i++, d++, p += 3) {
+            uint8_t r = *d & 0x00f8;
+            uint8_t g = (*d & 0x0007) << 5 | (*d & 0xe000) >> 11;
+            uint8_t b = (*d & 0x1f00) >> 5;
+            p[0] = r | r >> 5;
+            p[1] = g | g >> 6;
+            p[2] = b | b >> 5;
+        }
+    } else {
+        for (int i = 0; i < size; i++, d++, p += 3) {
+            uint8_t r = (*d & 0xf800) >> 8;
+            uint8_t g = (*d & 0x07e0) >> 3;
+            uint8_t b = (*d & 0x001f) << 3;
+            p[0] = r | r >> 5;
+            p[1] = g | g >> 6;
+            p[2] = b | b >> 5;
+        }
+    }
+}
diff --git a/ext/decoders/native/rgb.h b/ext/decoders/native/rgb.h
new file mode 100644
index 0000000..d144634
--- /dev/null
+++ b/ext/decoders/native/rgb.h
@@ -0,0 +1,8 @@
+#ifndef RGB_H
+#define RGB_H
+
+#include <stdint.h>
+
+void decode_rgb565(const uint16_t*, const int, const int, uint8_t*);
+
+#endif /* end of include guard: RGB_H */
diff --git a/lib/mikunyan/decoders/image_decoder.rb b/lib/mikunyan/decoders/image_decoder.rb
index f551729..eab79ec 100644
--- a/lib/mikunyan/decoders/image_decoder.rb
+++ b/lib/mikunyan/decoders/image_decoder.rb
@@ -127,15 +127,7 @@ module Mikunyan
         # @param [Symbol] endian endianness of binary
         # @return [ChunkyPNG::Image] decoded image
         def self.decode_rgb565(width, height, bin, endian = :big)
-            mem = String.new(capacity: width * height * 3)
-            (width * height).times do |i|
-                c = endian == :little ? BinUtils.get_int16_le(bin, i*2) : BinUtils.get_int16_be(bin, i*2)
-                r = (c & 0xf800) >> 8
-                g = (c & 0x07e0) >> 3
-                b = (c & 0x001f) << 3
-                BinUtils.append_int8!(mem, r | r >> 5, g | g >> 6, b | b >> 5)
-            end
-            ChunkyPNG::Image.from_rgb_stream(width, height, mem).flip
+            ChunkyPNG::Image.from_rgb_stream(width, height, DecodeHelper.decode_rgb565(bin, width * height, endian == :big)).flip
         end
 
         # Decode image from A8 binary