Further　compression　of　encoded　video　bitstream　is　necessary　to　save　storage　space　and　bandwidth.　However,　video　bitstream　is　a　compressed　version　of　raw　data　with　video　encoders　according　to　different　standards,　and　hence　the　signal　redundancy　is　already　reduced　compared　with　original　video　data.　Recompression　of　video　stream　requires　further　exploring　the　correlation　remained.　Transform　coding　as　a　part　of　hybrid　video　coding　framework　adopted　in　the　latest　video　coding　standards　such　as　discrete　cosine　transform　(DCT)　decorrelates　predictive　residual　signal　for　efficient　quantization　and　entropy　coding.　Nevertheless,　considerable　amount　of　statistical　correlation　still　remains　in　the　transform　coefficients　that　further　reducing　the　redundancy　can　lead　to　improved　coding　efficiency.　In　this　work,　we　propose　a　video　stream　recompression　scheme　based　on　further　sparse　representation　of　DCT　coefficients.　Dictionary-based　sparse　representation　method　is　used　after　DCT　coefficients　are　obtained　as　a　secondary　transform　module.　Moreover,　the　proposed　scheme　leverages　the　property　of　DPCM　and　avoids　sending　bits　of　dictionary　by　forming　redundant　dictionaries　from　DCT　coefficients　of　previously　decoded　frames.　Experimental　results　demonstrate　that　the　proposed　recompression　framework　further　reduces　the　bitrate　of　original　H.264　bitstream　by　more　than　60%　while　maintains　similar　subjective　quality.　©　2023　Copyright　held　by　the　owner/author(s).