In　this　paper,　a　novel　blocking　artifacts　reduction　algorithm　for　low-bit-rate　compressed　images　is　proposed.　The　proposed　algorithm　adaptively　adjusts　Discrete　Cosine　Transform　(DCT)　coefficients　of　the　shifted　blocks　to　remove　the　significant　blocking　artifacts.　Then,　the　bilateral　filter　is　introduced　to　smooth　the　subject　DCT　blocks.　The　subject　DCT　block　is　predicted　via　the　weighted　average　of　overlapping　neighboring　blocks　in　DCT　domain.　The　weights　are　determined　according　to　both　the　similarity　between　the　subject　coded　block　and　the　spatial　distance　of　the　overlapped　neighboring　blocks.　Image　fidelity　has　been　further　improved　via　a　quantization　constraint.　Experimental　results　illustrated　that　the　proposed　method　achieves　better　results　in　both　subjective　and　objective　quality.